JP4713278B2 - Polycrystalline semiconductor wafer visual inspection method and visual inspection apparatus - Google Patents

Description

  The present invention relates to a polycrystalline semiconductor wafer appearance inspection method and an appearance inspection apparatus for inspecting the appearance of a polycrystalline semiconductor wafer.

  A polycrystalline semiconductor wafer such as a polycrystalline silicon wafer is formed by cutting a silicon ingot into a certain thickness, for example, 200 μm, using a wire saw. In the wire saw, a workpiece, for example, a silicon ingot is linearly moved and pressed against wire rows of a constant pitch parallel to each other, and cut. At this time, the workpiece is cut by a lapping action to form a wafer by supplying a machining fluid containing abrasive grains between the workpiece and the wire while feeding the wire in the linear direction. This wire saw has an advantage that a large number of wafers having a constant thickness can be obtained simultaneously.

  The cut wafer is transferred to the chemical solution cleaning step and cleaned, but there are cases where dirt remains after cleaning. Since the characteristics of a solar cell manufactured using a dirty wafer are deteriorated, the dirty wafer cannot be passed through the cell manufacturing process. Therefore, in order to prevent the wafer from flowing into the cell manufacturing process, a wafer appearance inspection is performed to check whether dirt is attached. Actually, the wafer appearance inspection relies on the sensory inspection of the worker, that is, visual inspection. In recent years, silicon for solar cells has been insufficient, and the number of wafers has been increased by making the wafer thinner. However, in the wafer appearance inspection, manual handling is becoming difficult as the wafer is made thinner.

As a first conventional technique for performing an appearance inspection by image processing without relying on human hands, there is a solar cell inspection apparatus that determines the presence or absence of crack defects on the surface of a solar battery cell. This solar cell inspection device switches between oblique illumination and epi-illumination, and uses ITV (Industrial
Television) The existence of crack defects on the surface of the solar battery cell is determined based on the difference in the image of the surface of the solar battery cell taken by the camera (see, for example, Patent Document 1).

  As a second conventional technique, there is a substrate appearance inspection apparatus for inspecting the appearance of a substrate used for manufacturing a semiconductor device or a display device. This substrate appearance inspection apparatus captures an image of a surface of an inspection object such as a semiconductor device mounted on a stage or a substrate used for manufacturing a display device with an oblique illumination on the inspection object, and captures the image. Is detected using a so-called image recognition technique to detect foreign matter adhering to the surface. Further, the quality of the object to be inspected is judged by comparing the number of foreign matters adhering to the surface with a predetermined reference number. The oblique illumination is tilted by 10 to 30 degrees with respect to the surface of the inspection object and illuminated from obliquely above, so that foreign matters can be projected more clearly (see, for example, Patent Document 2).

JP-A-3-218045 JP 2001-33395 A

  In the first conventional technique, in the case of epi-illumination, since the crack defect does not appear in the image captured by the ITV camera, the crack defect can be recognized as a difference from the image captured by the oblique illumination. However, both the stain and the grain boundary of the polycrystalline silicon wafer also appear in the image captured by the epi-illumination, so that there is a problem that the stain and the grain boundary cannot be distinguished.

  In the second conventional technique, the surface of a single crystal silicon wafer for a semiconductor device and a glass plate for a display device are uniform and have no pattern such as a grain boundary. Therefore, based on an image captured by oblique illumination. Thus, foreign matters such as dirt can be detected. However, since the polycrystalline silicon wafer has a pattern due to grain boundaries, there is a problem that the grain boundaries and dirt cannot be distinguished from an image captured by oblique illumination.

  An object of the present invention is to provide an appearance inspection method and an appearance inspection apparatus capable of detecting dirt even in a polycrystalline semiconductor wafer having a grain boundary.

The present invention includes an imaging step of individually imaging two opposing surfaces of a polycrystalline semiconductor wafer,
The respective image data based on the images of the two surfaces captured in the imaging process are compared for each pixel at a position where the two surfaces are opposed to each other, and mismatch image data representing an image having a mismatch pixel as a mismatch pixel is obtained. A comparison process to generate ;
Judgment to determine that one of the two surfaces is soiled when a region of the unmatched pixels among the pixels of the image indicated by the unmatched image data generated in the comparison step satisfies a predetermined stain condition and a step seen including,
The predetermined dirt condition is a second region in which a region composed of non-matching pixels is a region composed of pixels equal to or more than a first predetermined number of pixels, and a vertical pixel and a horizontal pixel are respectively predetermined. A method for inspecting the appearance of a polycrystalline semiconductor wafer, characterized in that the regions are adjacent to each other by at least the number of pixels .

According to the present invention, in the imaging step, the two opposing surfaces of the polycrystalline semiconductor wafer are individually imaged, and in the comparison step, the respective image data based on the images of the two surfaces imaged in the imaging step, Compared for each pixel at a position where the two surfaces face each other, non-matching image data representing an image having a non- matching pixel as a non- matching pixel is generated . In the determination step, the image indicated by the non- matching image data generated in the comparison step When a region consisting of non-matching pixels satisfies a predetermined dirt condition, it is determined that one of the two faces is dirty. Further, the predetermined dirt condition is that the region composed of the mismatched pixels is a region composed of pixels equal to or more than the first predetermined number of pixels, and the vertical pixel and the horizontal pixel are predetermined. 2 is an area adjacent to at least two pixels.

  As described above, since the respective image data based on the images obtained by imaging the two surfaces of the polycrystalline semiconductor wafer are compared, the images appearing at the positions where both surfaces are opposed to each other are canceled, and only the image appearing on any one surface is detected. Can be extracted.

In the present invention, the image data is dirt candidate image data indicating, for each pixel, whether or not the pixel data is a candidate for a pixel representing dirt.
In the comparison step, the stain candidate image data of the two surfaces is compared for each pixel at a position where the two surfaces are opposed to each other, and data indicating whether or not the pixel candidate represents a stain is opposed to each other. the pixel position of the polycrystalline semiconductor wafer corresponding to the position of the pixel, the mismatch image data to mismatch pixels characterized Rukoto forming raw.

According to the present invention, the stain candidate image data of the two surfaces are compared for each pixel at the position where the two surfaces face each other, and the data indicating whether or not the pixel candidate represents the stain is opposed to each other. the pixel position of the polycrystalline semiconductor wafer corresponding to the position of the pixel, as they may mismatch image data constituting raw and mismatched pixels, polycrystalline semiconductor wafer of the grain boundary of the partial images indicated by the dirty candidate image data of two planes Even if it appears, the grain boundary part is canceled out and does not become a mismatched pixel, but only the dirty part can be extracted as a mismatched pixel, and the region consisting of the extracted mismatched pixels satisfies a predetermined dirt condition It can be determined that there is dirt.

The present invention further includes a generation step of generating the stain candidate image data,
In the imaging step, each surface is sequentially irradiated with light from a plurality of angles, and images of each surface are captured for each angle at which the light is irradiated.
In the generation step, the brightness difference is calculated for each pixel at the same position of the polycrystalline semiconductor wafer, and the calculated brightness difference is smaller than the predetermined brightness difference. Image data of a dirt candidate image having a pixel at a position as a difference as the pixel candidate is generated for each surface as the dirt candidate image data.

  According to the present invention, each surface is sequentially irradiated with light from a plurality of angles, an image of each surface is captured for each angle at which light is irradiated, and image data captured from each of the plurality of angles is obtained. A brightness difference is calculated for each pixel at the same position on the polycrystalline semiconductor wafer, and a stain candidate image is obtained with a pixel at a position where the calculated brightness difference is smaller than a predetermined brightness difference as a candidate for a pixel representing dirt. Since the stain candidate image data to be represented is generated for each surface, it is possible to generate the stain candidate image data obtained by extracting the portion where the brightness difference of the image data is small.

The present invention further includes a generation step of generating the stain candidate image data,
In the imaging step, each surface is sequentially irradiated with light from a plurality of angles, and images of each surface are captured for each angle at which the light is irradiated.
In the generation step, the brightness difference is calculated for each pixel at the same position of the polycrystalline semiconductor wafer, and the calculated brightness difference is smaller than the predetermined brightness difference. Image data of a dirt candidate contour image having a pixel at the outermost edge of the area as a pixel candidate out of areas composed of pixels at positions that are differences is generated as the dirt candidate image data for each surface. To do.

  According to the present invention, each surface is sequentially irradiated with light from a plurality of angles, an image of each surface is captured for each angle at which light is irradiated, and image data captured from each of the plurality of angles is obtained. The brightness difference is calculated for each pixel at the same position on the polycrystalline semiconductor wafer, and the pixel at the outermost edge of the area is composed of pixels at positions where the calculated brightness difference is smaller than the predetermined brightness difference. Is generated for each surface, and the candidate stain image data obtained by extracting the contour of the portion where the brightness difference of the image data is small is generated. it can.

According to another aspect of the present invention, there are provided imaging means for individually imaging two opposing surfaces of a polycrystalline semiconductor wafer,
The respective image data based on the images of the two surfaces captured by the imaging means are compared for each pixel at a position where the two surfaces are opposed to each other, and mismatch image data representing an image having a mismatch pixel as a mismatch pixel is obtained. A comparison means to generate ;
Judgment to determine that one of the two surfaces is soiled when a region of the unmatched pixels among the pixels of the image indicated by the unmatched image data generated by the comparing means satisfies a predetermined stain condition and it means only including,
The predetermined dirt condition is a second region in which a region composed of non-matching pixels is a region composed of pixels equal to or more than a first predetermined number of pixels, and a vertical pixel and a horizontal pixel are respectively predetermined. An appearance inspection apparatus for a polycrystalline semiconductor wafer, characterized in that it is a region adjacent to at least the number of pixels .

According to the present invention, two opposing surfaces of the polycrystalline semiconductor wafer are individually imaged by the imaging unit, and each image data based on the images of the two surfaces imaged by the imaging unit by the comparison unit is two faces and are compared for each pixel at a position opposing mismatch image data representing an image to be inconsistent pixel pixel is disagreement is generated, the determination means, an image representing a mismatch image data generated by the comparison means When a region consisting of non-matching pixels satisfies a predetermined dirt condition, it is determined that one of the two faces is dirty. Further, the predetermined dirt condition is that the region composed of the mismatched pixels is a region composed of pixels equal to or more than the first predetermined number of pixels, and the vertical pixel and the horizontal pixel are predetermined. 2 is an area adjacent to at least two pixels.

  In this way, since the respective image data based on the images obtained by imaging two parallel surfaces of the polycrystalline semiconductor wafer are compared, the images appearing at the positions where both surfaces face each other are canceled out and appear on any one surface. Only images can be extracted.

In the present invention, the image data is dirt candidate image data indicating, for each pixel, whether or not the pixel data is a candidate for a pixel representing dirt.
The comparison means compares the stain candidate image data of the two surfaces for each pixel at a position where the two surfaces are opposed to each other, and the data indicating whether or not the pixel candidates indicate the stain are opposed to each other. the pixel position of the polycrystalline semiconductor wafer corresponding to the position of the pixel, the mismatch image data to mismatch pixels characterized Rukoto forming raw.

According to the present invention, the stain candidate image data of the two surfaces are compared for each pixel at the position where the two surfaces face each other, and the data indicating whether or not the pixel candidate represents the stain is opposed to each other. the pixel position of the polycrystalline semiconductor wafer corresponding to the position of the pixel, as they may mismatch image data constituting raw and mismatched pixels, polycrystalline semiconductor wafer of the grain boundary of the partial images indicated by the dirty candidate image data of two planes Even if it appears, the grain boundary part is canceled out and does not become a mismatched pixel, but only the dirty part can be extracted as a mismatched pixel, and the region consisting of the extracted mismatched pixels satisfies a predetermined dirt condition It can be determined that there is dirt.

Further, the present invention provides a generating means for generating the stain candidate image data,
Irradiation means for irradiating light from a plurality of angles to the two surfaces,
The imaging means captures an image of each surface for each angle at which light is irradiated by the irradiation means,
The generation unit calculates a brightness difference for each pixel at the same position on the polycrystalline semiconductor wafer, and the calculated brightness difference is smaller than a predetermined brightness difference. Image data of a dirt candidate image having a pixel at a position as a difference as the pixel candidate is generated for each surface as the dirt candidate image data.

  According to the present invention, each surface is sequentially irradiated with light from a plurality of angles, an image of each surface is captured for each angle at which light is irradiated, and image data captured from each of the plurality of angles is obtained. A brightness difference is calculated for each pixel at the same position on the polycrystalline semiconductor wafer, and a stain candidate image is obtained with a pixel at a position where the calculated brightness difference is smaller than a predetermined brightness difference as a candidate for a pixel representing dirt. Since the stain candidate image data to be represented is generated for each surface, it is possible to generate the stain candidate image data obtained by extracting the portion where the brightness difference of the image data is small.

Further, the present invention provides a generating means for generating the stain candidate image data,
Irradiation means for irradiating light from a plurality of angles to the two surfaces,
The imaging means captures an image of each surface for each angle at which light is irradiated by the irradiation means,
The generation unit calculates a brightness difference for each pixel at the same position on the polycrystalline semiconductor wafer, and the calculated brightness difference is smaller than a predetermined brightness difference. Image data of a dirt candidate contour image having a pixel at the outermost edge of the area as a pixel candidate out of areas composed of pixels at positions that are differences is generated as the dirt candidate image data for each surface. To do.

  According to the present invention, each surface is sequentially irradiated with light from a plurality of angles, an image of each surface is captured for each angle at which light is irradiated, and image data captured from each of the plurality of angles is obtained. The brightness difference is calculated for each pixel at the same position on the polycrystalline semiconductor wafer, and the pixel at the outermost edge of the area is composed of pixels at positions where the calculated brightness difference is smaller than the predetermined brightness difference. Is generated for each surface, and the candidate stain image data obtained by extracting the contour of the portion where the brightness difference of the image data is small is generated. it can.

  According to the present invention, it is possible to cancel an image appearing at a position where both faces face each other and extract only an image appearing on any one face. Therefore, even in a polycrystalline semiconductor wafer having a grain boundary, It is possible to cancel the pattern due to the appearing grain boundary and to detect the dirt adhering to any one surface.

  Furthermore, since this appearance inspection method can detect dirt adhering to any one of the two surfaces, it replaces the sensory inspection, that is, the visual inspection that has been performed manually, and the appearance inspection of the polycrystalline semiconductor wafer. Can be automated. Further, since it is possible to determine the presence / absence of dirt even on a thin polycrystalline semiconductor wafer that cannot be handled manually, a wafer without dirt can be provided to the cell manufacturing process of the solar cell.

  Further, according to the present invention, even if the grain boundary portion of the polycrystalline semiconductor wafer appears in the image indicated by the two surface stain candidate image data, the grain boundary portion is canceled and does not become a mismatched pixel, Can be extracted as non-matching pixels, and when a region consisting of the extracted non-matching pixels satisfies a predetermined dirt condition, it can be determined that there is dirt. Therefore, even in a polycrystalline semiconductor wafer having a grain boundary, it is possible to cancel the pattern due to the grain boundary appearing on both surfaces, and to detect the dirt adhering to any one surface.

  Further, according to the present invention, it is possible to generate dirt candidate image data obtained by extracting a part having a small brightness difference in the image data, and therefore it is possible to extract only a grain boundary and a dirt part having a small brightness difference.

  Further, according to the present invention, it is possible to generate the stain candidate image data in which the contour of the portion where the brightness difference of the image data is small is extracted, so that the contour of the grain boundary and the stain portion where the brightness difference is small can be extracted. . Since it is determined whether or not a predetermined dirt condition is satisfied by the contour, the determination can be made with a smaller amount of information than when determining by the region.

  Further, according to the present invention, since the images appearing at the positions where both surfaces face each other can be canceled and only the image appearing on any one surface can be extracted, both surfaces of a polycrystalline semiconductor wafer having a grain boundary can be obtained. It is possible to cancel the pattern due to the grain boundary appearing on the surface, and to detect the dirt adhering to any one surface.

  Furthermore, since this appearance inspection apparatus can detect dirt adhering to any one of the two surfaces, it replaces the sensory inspection, that is, visual inspection, which has been performed manually, and the appearance inspection of the polycrystalline semiconductor wafer. Can be automated. Further, since it is possible to determine the presence / absence of dirt even on a thin polycrystalline semiconductor wafer that cannot be handled manually, a wafer without dirt can be provided to the cell manufacturing process of the solar cell.

  Further, according to the present invention, even if the grain boundary portion of the polycrystalline semiconductor wafer appears in the image indicated by the two surface stain candidate image data, the grain boundary portion is canceled and does not become a mismatched pixel, Can be extracted as non-matching pixels, and when a region consisting of the extracted non-matching pixels satisfies a predetermined dirt condition, it can be determined that there is dirt. Therefore, even in a polycrystalline semiconductor wafer having a grain boundary, it is possible to cancel the pattern due to the grain boundary appearing on both surfaces, and to detect the dirt adhering to any one surface.

  Further, according to the present invention, it is possible to generate dirt candidate image data obtained by extracting a part having a small brightness difference in the image data, and therefore it is possible to extract only a grain boundary and a dirt part having a small brightness difference.

  Further, according to the present invention, it is possible to generate the stain candidate image data in which the contour of the portion where the brightness difference of the image data is small is extracted, so that the contour of the grain boundary and the stain portion where the brightness difference is small can be extracted. . Since it is determined whether or not a predetermined dirt condition is satisfied by the contour, the determination can be made with a smaller amount of information than when determining by the region.

  FIG. 1 schematically shows a configuration of a polycrystalline silicon wafer visual inspection apparatus 1 according to an embodiment of the present invention. A polycrystalline silicon wafer visual inspection apparatus 1 which is an visual inspection apparatus is an visual inspection apparatus that detects contamination of a polycrystalline semiconductor wafer, for example, a polycrystalline silicon wafer 2, and includes a computer 10, a surface imaging device 20, and a polycrystalline silicon wafer inversion device. 30, a back surface imaging device 40, and a transport device 50.

  The polycrystalline silicon wafer 2 is formed by cutting a silicon ingot into a certain thickness, for example, 200 μm, using a wire saw, and has two surfaces formed by cutting. Hereinafter, one of these surfaces is referred to as a front surface (hereinafter referred to as a front surface), and the other surface is referred to as a back surface.

  The computer 10 is configured by a personal computer including an output device such as a monitor and an input device such as a keyboard. The computer 10 further includes a storage device such as a CPU (Central Processing Unit), a semiconductor memory, and a hard disk, and includes a front surface imaging device 20, a polycrystalline silicon wafer reversing device 30, a back surface imaging device 40, and a transport device. 50 is controlled.

  The surface imaging device 20 is an imaging device that images the surface of the polycrystalline silicon wafer 2, and includes a light source unit 21, a camera 22, a diffusion plate 23, and a bridge pier 24. The light source unit 21 includes four parts, and each part includes two light sources. The four portions of the light source unit 21 are installed so as to surround the polycrystalline silicon wafer 2 from four sides. Each light source is configured by, for example, a fluorescent tube, and one of the two light sources included in each portion irradiates light on the surface of the polycrystalline silicon wafer 2 at an angle of 10 degrees to 30 degrees. The other light sources are arranged to irradiate light at an angle of 60 degrees to 80 degrees with respect to the surface of the polycrystalline silicon wafer 2.

  The camera 22 is configured by a digital still camera including, for example, a 1 million pixel CCD (Charge Coupled Device), and is installed on the bridge pier 24 so as to face the surface of the polycrystalline silicon wafer 2. The camera 22 images the surface of the polycrystalline silicon wafer 2 according to an instruction from the computer 10, and image data of the captured image is transferred to the computer 10. The diffusion plate 23 is for diffusing the light emitted from the light source unit 21, and is installed between the light source unit 21 and the polycrystalline silicon wafer 2.

  The polycrystalline silicon wafer reversing device 30 includes an air chuck, that is, an air suction device 31 and a rotating rod 32. The air suction device 31 sucks the polycrystalline silicon wafer 2 by sucking air. The rotary rod 32 can rotate the air suction device 31 attached to the rotary rod 32 by 180 degrees with the rotary rod 32 as the rotation axis, and the polycrystalline silicon wafer 2 adsorbed by the air suction device 31 is turned over. can do. That is, the surface of the polycrystalline silicon wafer 2 facing upward can be directed downward and the back surface can be directed upward.

  The back surface imaging device 40 is an imaging device that images the back surface of the polycrystalline silicon wafer 2, and includes a light source unit 41, a camera 42, a diffusion plate 43, and a bridge pier 44. The light source unit 41 includes four parts, and each part includes two light sources. The four portions of the light source unit 41 are installed so as to surround the polycrystalline silicon wafer 2 from four sides. Each light source is configured by, for example, a fluorescent tube, and one of the two light sources included in each portion irradiates light at an angle of 10 degrees to 30 degrees with respect to the back surface of the polycrystalline silicon wafer 2. The other light sources are arranged so as to irradiate light at an angle of 60 degrees to 80 degrees with respect to the back surface of the polycrystalline silicon wafer 2.

  The camera 42 is configured by a digital still camera having a CCD with 1 million pixels, for example, and is installed on the bridge pier 44 so as to face the back surface of the polycrystalline silicon wafer 2. The camera 42 images the back surface of the polycrystalline silicon wafer 2 according to an instruction from the computer 10, and image data of the captured image is transferred to the computer 10. The diffusion plate 43 is for diffusing the light emitted from the light source unit 41, and is installed between the light source unit 41 and the polycrystalline silicon wafer 2.

  The transfer device 50 is a transfer device that transfers the polycrystalline silicon wafer 2, and includes two rails that support the polycrystalline silicon wafer 2 and a stage on which the polycrystalline silicon wafer 2 is placed and moved or stopped. . The stage is not shown because it is smaller than the polycrystalline silicon wafer 2. The polycrystalline silicon wafer 2 that has been imaged by the front surface imaging device 20 is transported to the polycrystalline silicon wafer reversing device 30, and the polycrystalline silicon wafer 2 that is turned upside down by the polycrystalline silicon wafer reversing device 30 is transferred to the back surface imaging device 40. Transport to.

  The computer 10 changes the angle of the light emitted from the light source unit 21 and takes two types of images of the surface of the polycrystalline silicon wafer 2 with the surface imaging device 20. That is, a low-angle image taken with illumination that irradiates light at an angle of 10 to 30 degrees with respect to the surface of the polycrystalline silicon wafer 2, and 60 to 80 degrees with respect to the surface of the polycrystalline silicon wafer 2. A high-angle image is captured with illumination that emits light at an angle. Image data of two types of images captured by the surface imaging device 20 is transferred to the computer 10 and stored in, for example, a hard disk.

  Next, the computer 10 changes the angle of the light emitted from the light source unit 41 and takes two types of images of the back surface of the polycrystalline silicon wafer 2 by the back surface imaging device 40. That is, a low angle image and a high angle image with respect to the back surface of the polycrystalline silicon wafer 2 are taken. Image data of two types of images captured by the back surface imaging device 40 is transferred to the computer 10 and stored in, for example, a hard disk.

  Further, the computer 10 uses the two types of image data obtained by imaging the surface of the polycrystalline silicon wafer 2 to obtain the surface contamination candidate image data and the two types of image data obtained by imaging the back surface of the polycrystalline silicon wafer 2. After the image data is generated, the presence / absence of dirt attached to the front surface or the back surface is determined by comparing the generated front surface dirt candidate image data and the back surface dirt candidate image data. The stain candidate image data is image data of a stain candidate image indicating, for each pixel, pixel candidates representing the stain attached to the front surface or the back surface. In the dirt candidate image, dirt and grain boundary portions are extracted.

  Since the polycrystalline silicon wafer 2 is as thin as, for example, 200 μm, the pattern due to the grain boundary on the front surface and the pattern due to the grain boundary on the back surface are the same if they are reversed upside down or left and right. That is, even if a pattern due to the front grain boundary appears in the image indicated by the front dirt candidate image data and a pattern due to the rear grain boundary appears in the image indicated by the rear dirt candidate image data, By taking the exclusive OR of the stain candidate image data on the back surface for each pixel at the position where the two surfaces face each other, the pattern due to the grain boundary on the front surface and the pattern due to the grain boundary on the back surface can be offset, Only the dirty portion remains as a difference, and if there is a difference, it can be determined that there is a stain.

  The imaging unit is, for example, the camera 22 and the camera 42, the comparison unit, the determination unit, and the generation unit are, for example, the computer 10, and the irradiation unit is, for example, the light source unit 21 and the light source unit 41.

  The polycrystalline silicon wafer 2 is, for example, a 125 mm square and 200 μm thick polycrystalline silicon wafer. The following is an appearance inspection performed by the polycrystalline silicon wafer visual inspection apparatus 1 on the polycrystalline silicon wafer 2 with the dirt on the surface. The embodiment will be described in detail with reference to FIG.

  The polycrystalline silicon wafer 2 transported by the transport device 50 stops at the position of the surface imaging device 20 in order to image the surface. When the light source unit 21 irradiates the surface of the polycrystalline silicon wafer 2 with light at an angle of 15 degrees, for example, the surface imaging device 20 images the surface of the polycrystalline silicon wafer 2 and image data of the captured image. To the computer 10. The computer 10 stores the received image data in the hard disk as image data of a low-angle image of the surface. The camera 22 is installed so that, of the four sides of the polycrystalline silicon wafer 2, the side in the transfer direction of the transfer device 50 is reflected on the upper part of the monitor of the computer 10.

  Next, when the light source unit 21 irradiates the surface of the polycrystalline silicon wafer 2 with light at an angle of, for example, 75 degrees, the surface imaging device 20 images the surface of the polycrystalline silicon wafer 2 and the captured image. Are sent to the computer 10. The computer 10 stores the received image data in the hard disk as image data of a high-angle image of the surface.

  FIG. 2 shows an example of a low-angle image of the surface of the polycrystalline silicon wafer 2 imaged by the surface imaging device 20 shown in FIG. A pattern and dust due to grain boundaries of the polycrystalline silicon wafer 2 are shown.

  FIG. 3 shows an example of a high-angle image of the surface of the polycrystalline silicon wafer 2 imaged by the surface imaging device 20 shown in FIG. The high-angle image shown in FIG. 3 shows patterns and dust due to the grain boundaries of the polycrystalline silicon wafer 2 as in the low-angle image shown in FIG. Compared to the low-angle image shown in FIG.

  Referring to FIG. 1, when imaging with surface imaging device 20 is completed, conveyance device 50 conveys polycrystalline silicon wafer 2 on which imaging of the surface has been completed to polycrystalline silicon wafer reversing device 30. The polycrystalline silicon wafer reversing device 30 adsorbs the polycrystalline silicon wafer 2 transported by the transport device 50 with the air suction device 31 and rotates the rotating rod 32 so that the back surface of the polycrystalline silicon wafer 2 faces upward. Invert to face. The air suction device 31 releases the adsorption, and places the polycrystalline silicon wafer 2 that is inverted and the back surface is directed upward on the transfer device 50. The transport device 50 transports the inverted polycrystalline silicon wafer 2 to the back surface imaging device 40.

  The polycrystalline silicon wafer 2 transported by the transport device 50 stops at the position of the back surface imaging device 40 in order to image the back surface. When the light source unit 41 irradiates light on the back surface of the polycrystalline silicon wafer 2 at an angle of, for example, 15 degrees, the back surface imaging device 40 captures the back surface of the polycrystalline silicon wafer 2 and image data of the captured image. To the computer 10. The computer 10 stores the received image data in the hard disk as image data of a low-angle image of the back surface. The camera 42 is installed so that, of the four sides of the polycrystalline silicon wafer 2, the side in the transfer direction of the transfer device 50 is reflected on the upper part of the monitor of the computer 10.

  Next, when the light source unit 41 irradiates the back surface of the polycrystalline silicon wafer 2 with light at an angle of, for example, 75 degrees, the back surface imaging device 40 images the back surface of the polycrystalline silicon wafer 2 and the captured image. Are sent to the computer 10. The computer 10 stores the received image data in the hard disk as image data of a high-angle image on the back surface. When the imaging by the back surface imaging device 40 is completed, the transport device 50 transports the polycrystalline silicon wafer 2 whose back surface has been imaged to the next step.

  First, the computer 10 calculates the brightness difference for each pixel at the same position on the polycrystalline silicon wafer 2 with respect to the image data of the low-angle image of the surface and the image data of the high-angle image of the surface stored in the hard disk. Next, stain candidate image data representing a stain candidate image in which a pixel at a position where the calculated brightness difference is smaller than a predetermined brightness difference is a candidate for a pixel representing stain is generated.

  Further, the image data of the back side low angle image and the back side high angle image stored in the hard disk is similarly calculated by calculating the brightness difference for each pixel at the same position of the polycrystalline silicon wafer 2. Further, stain candidate image data representing a stain candidate image, in which a pixel at a position where the brightness difference is smaller than a predetermined brightness difference is a candidate for a pixel representing stain, is generated. Since the back surface of the polycrystalline silicon wafer 2 is imaged after the polycrystalline silicon wafer 2 is inverted 180 degrees by the polycrystalline silicon wafer reversing device 30, the back surface of the polycrystalline silicon wafer 2 is vertically Inverted. Therefore, in order to match the position of each pixel of the stain candidate image indicated by the stain candidate image data on the back surface with the position of each pixel of the stain candidate image indicated by the stain candidate image data on the front surface, the computer 10 The image data is converted into inverted image data representing a dirt candidate image that is inverted upside down.

  FIG. 4 shows an example of a stain candidate image on the surface of the polycrystalline silicon wafer 2. FIG. 4 shows a stain candidate image generated based on the low-angle image of the surface and the high-angle image of the surface. In this embodiment, the stain is attached to the surface of the polycrystalline silicon wafer 2. In addition, the grain boundary and the part of dirt are shown as dirt candidate areas. In FIG. 4, a dirt candidate area 71 due to dirt is shown in the area 61.

  FIG. 5 shows an example of an inverted image obtained by vertically inverting the dirt candidate image on the back surface of the polycrystalline silicon wafer 2. FIG. 5 shows a reverse image of the back surface indicated by the reverse image data. In this embodiment, since no dirt is attached to the back surface, only the grain boundary portion is shown as a stain candidate region. The inverted image data is image data representing a dirt candidate image obtained by inverting the dirt candidate image data generated based on the image data of the low angle image and the image data of the high angle image of the back surface of the polycrystalline silicon wafer 2 upside down. It is converted. There is no dirt candidate area in the area 62 shown in FIG.

  The computer 10 compares the stain candidate image data on the front surface and the reverse image data on the back surface for each pixel at the same position, that is, at a position where the two surfaces face each other, and indicates whether or not the pixel candidate represents a stain. Disagreement image data representing an image in which the pixel at the position of the polycrystalline silicon wafer 2 corresponding to the position of the opposing pixel having a mismatch is used as the mismatch pixel is generated. Then, it is determined that there is dirt when a region composed of mismatched pixels among the pixels of the image indicated by the mismatched image data satisfies a predetermined dirt condition.

  In the computer 10, the region satisfying the predetermined stain condition is a region corresponding to the stain candidate region included in the stain candidate image on the front surface or a region corresponding to the stain candidate region included in the stain candidate image on the back surface. Determine whether. If the area corresponds to the dirt candidate area of the dirt candidate image on the front surface, it is specified that dirt is attached to the surface of the polycrystalline silicon wafer 2, or the area is the candidate dirt image on the back face. If it is an area corresponding to the dirt candidate area, it can be specified that dirt is attached to the back surface of the polycrystalline silicon wafer 2.

  FIG. 6 shows an image obtained by enlarging the region 61 in the dirt candidate image shown in FIG. A region 61 is a part of a stain candidate image on the surface of the polycrystalline silicon wafer 2. The area 61 is an area composed of 8 pixels in the vertical and horizontal directions. One cell indicates one pixel, black pixels are dirt candidate pixels, and white pixels are non-dirty pixels. In the area 61, a dirt candidate area 71 due to dirt is shown. The dirt candidate area is an area due to dirt or grain boundaries.

  FIG. 7 shows an image in which the region 62 in the dirt candidate image shown in FIG. 5 is enlarged. The region 62 is a part of the reverse image of the back surface of the polycrystalline silicon wafer 2 and corresponds to the region 61 shown in FIG. Similar to the area 61, the area 62 is an area composed of 8 pixels in each of the vertical and horizontal directions. One cell indicates one pixel, a black pixel is a stain candidate pixel, and a white pixel is a pixel that is not a stain candidate. . The region 62 does not include a dirt candidate region.

  FIG. 8 shows an image of mismatched image data obtained by comparing the region 61 shown in FIG. 6 with the region 62 shown in FIG. The mismatched image data is determined by the computer 10 comparing the surface dirt candidate image data with the inverted image data obtained by vertically inverting the dirt candidate image data on the back surface for each pixel at a position where the two surfaces face each other. This is image data representing an image in which a pixel at a position corresponding to the position of an opposing pixel in which the data indicating whether or not the pixel candidate represents a mismatch is a mismatched pixel. In other words, the pixel-by-pixel data of the front surface stain candidate image data and the pixel-by-pixel data of the reverse image data of the back surface are data obtained by exclusive ORing the pixels at the same position. FIG. 8 shows an image of the region 61 shown in FIG. 6, that is, a part of the front surface dirt candidate image shown in FIG. 4, and an image of the region 62 shown in FIG. 7, that is, the reverse image of the back side shown in FIG. Is an image obtained by taking an exclusive OR of the corresponding parts of each pixel at the same position.

  The predetermined dirt condition for determining that there is dirt is that the area composed of non-matching pixels is an area composed of 16 pixels or more in, for example, an area of 8 pixels in the vertical direction and 8 pixels in the horizontal direction, and in the vertical direction. This is a condition that the pixel and the pixel in the horizontal direction are areas adjacent to each other by 3 pixels or more.

  The region 81 made up of non-matching pixels shown in FIG. 8 is a region made up of 25 pixels within an area of 8 pixels in the vertical direction and 8 pixels in the horizontal direction, and each of the vertical and horizontal pixels is 3 pixels or more. It is an adjacent area. In the computer 10, the region 81 made up of mismatched pixels shown in FIG. 8 is a region composed of 16 or more pixels within a predetermined dirt condition, that is, within an area of 8 pixels in the vertical direction and 8 pixels in the horizontal direction, and Since the condition that the pixels in the vertical direction and the pixels in the horizontal direction are regions adjacent to each other by 3 pixels or more is satisfied, it is determined that the polycrystalline silicon wafer 2 is contaminated.

  The computer 10 specifies that the dirt is attached to the surface of the polycrystalline silicon wafer 2 because the area 81 including the non-matching pixels shown in FIG. 8 corresponds to the dirt candidate area 71 shown in FIG.

  In this way, since the respective image data based on the images obtained by imaging the two surfaces of the polycrystalline semiconductor wafer, for example, the polycrystalline silicon wafer 2, that is, the front surface and the back surface are compared, the images appearing at the positions where both surfaces face each other are canceled out. Only images that appear on any one of the surfaces can be extracted. Therefore, even in a polycrystalline semiconductor wafer having a grain boundary, it is possible to cancel the pattern due to the grain boundary appearing on both surfaces, and to detect the dirt adhering to any one surface.

  Furthermore, since the polycrystalline silicon wafer appearance inspection apparatus 1 which is an appearance inspection apparatus can detect dirt adhering to any one of the two surfaces, that is, the front surface or the back surface, the sensory inspection performed manually. That is, the visual inspection can be replaced and the appearance inspection of the polycrystalline semiconductor wafer such as the polycrystalline silicon wafer 2 can be automated. Further, since it is possible to determine the presence / absence of dirt even on a thin polycrystalline semiconductor wafer that cannot be handled manually, a wafer without dirt can be provided to the cell manufacturing process of the solar cell.

  Furthermore, by comparing the candidate stain image data of two surfaces, that is, the candidate stain image data on the front surface and the candidate stain image data on the back surface, for each pixel at a position where the two surfaces face each other, Inconsistency image data representing an image in which a pixel at a position of a polycrystalline semiconductor wafer corresponding to the position of an opposing pixel in which data indicating whether or not there is a mismatch, for example, a position of a polycrystalline silicon wafer 2, is a mismatched pixel is generated, and the mismatch is generated. When the pixel region satisfies the predetermined dirt condition, it is determined that there is dirt. Therefore, even if the grain boundary portion of the polycrystalline semiconductor wafer appears in the image indicated by the two candidate dirt image data, The part is canceled out and does not become a mismatched pixel, but only the dirty part can be extracted as a mismatched pixel, and the region composed of the extracted mismatched pixels satisfies a predetermined dirt condition. When, it can be determined that there is contamination. Therefore, even in a polycrystalline semiconductor wafer having a grain boundary, it is possible to cancel the pattern due to the grain boundary appearing on both surfaces, and to detect the dirt adhering to any one surface.

  Furthermore, light is sequentially irradiated from two angles with respect to each surface, for example, a low angle of 10 to 30 degrees and a high angle of 60 to 80 degrees with respect to the front or back surface, and the light is irradiated at each angle. An image of each surface is picked up, and image data picked up from two angles, for example, image data of a low angle image and a high angle image are obtained for each pixel at the same position of a polycrystalline semiconductor wafer, for example, a polycrystalline silicon wafer 2. A brightness difference is calculated, and stain candidate image data representing a stain candidate image is generated for each surface, with a pixel at a position where the calculated brightness difference is smaller than a predetermined brightness difference as a pixel candidate representing stain. Therefore, it is possible to generate dirt candidate image data obtained by extracting a portion where the brightness difference of the image data is small. Therefore, it is possible to extract only the grain boundary and the soiled portion where the brightness difference is small.

  The polycrystalline silicon wafer 2 is, for example, a 125 mm square polycrystalline silicon wafer. The following is another implementation of the visual inspection performed by the polycrystalline silicon wafer visual inspection apparatus 1 on the polycrystalline silicon wafer 2 having a surface contaminated. An example will be described in detail with reference to FIG.

  The polycrystalline silicon wafer 2 transported by the transport device 50 stops at the position of the surface imaging device 20 in order to image the surface. When the light source unit 21 irradiates the surface of the polycrystalline silicon wafer 2 with light at an angle of 15 degrees, for example, the surface imaging device 20 images the surface of the polycrystalline silicon wafer 2 and image data of the captured image. To the computer 10. The computer 10 stores the received image data in the hard disk as image data of a low-angle image of the surface. The camera 22 is installed so that, of the four sides of the polycrystalline silicon wafer 2, the side in the transfer direction of the transfer device 50 is reflected on the upper part of the monitor of the computer 10.

  Next, when the light source unit 21 irradiates the surface of the polycrystalline silicon wafer 2 with light at an angle of, for example, 75 degrees, the surface imaging device 20 images the surface of the polycrystalline silicon wafer 2 and the captured image. Are sent to the computer 10. The computer 10 stores the received image data in the hard disk as image data of a high-angle image of the surface.

  When imaging by the surface imaging device 20 is completed, the transport device 50 transports the polycrystalline silicon wafer 2 whose surface has been imaged to the polycrystalline silicon wafer reversing device 30. The polycrystalline silicon wafer reversing device 30 adsorbs the polycrystalline silicon wafer 2 transported by the transport device 50 with the air suction device 31 and rotates the rotating rod 32 so that the back surface of the polycrystalline silicon wafer 2 faces upward. Invert to face. The air suction device 31 releases the adsorption, and places the polycrystalline silicon wafer 2 that is inverted and the back surface is directed upward on the transfer device 50. The transport device 50 transports the inverted polycrystalline silicon wafer 2 to the back surface imaging device 40.

  The polycrystalline silicon wafer 2 transported by the transport device 50 stops at the position of the back surface imaging device 40 in order to image the back surface. When the light source unit 41 irradiates light on the back surface of the polycrystalline silicon wafer 2 at an angle of, for example, 15 degrees, the back surface imaging device 40 captures the back surface of the polycrystalline silicon wafer 2 and image data of the captured image. To the computer 10. The computer 10 stores the received image data in the hard disk as image data of a low-angle image of the back surface. The camera 42 is installed so that, of the four sides of the polycrystalline silicon wafer 2, the side in the transfer direction of the transfer device 50 is reflected on the upper part of the monitor of the computer 10.

  Next, when the light source unit 41 irradiates the back surface of the polycrystalline silicon wafer 2 with light at an angle of, for example, 75 degrees, the back surface imaging device 40 images the back surface of the polycrystalline silicon wafer 2 and the captured image. Are sent to the computer 10. The computer 10 stores the received image data in the hard disk as image data of a high-angle image on the back surface. When the imaging by the back surface imaging device 40 is completed, the transport device 50 transports the polycrystalline silicon wafer 2 whose back surface has been imaged to the next step.

  First, the computer 10 calculates the brightness difference for each pixel at the same position on the polycrystalline silicon wafer 2 with respect to the image data of the low-angle image of the surface and the image data of the high-angle image of the surface stored in the hard disk. Next, among the regions composed of pixels at positions where the calculated lightness difference is smaller than the predetermined lightness difference, the pixel at the outermost edge of the region, that is, the pixel forming the contour of the region, is a pixel candidate that represents a stain. The dirt candidate image data representing the dirt candidate contour image to be, that is, the dirt candidate contour image data is generated.

  Further, the image data of the back side low angle image and the back side high angle image stored in the hard disk is similarly calculated by calculating the brightness difference for each pixel at the same position of the polycrystalline silicon wafer 2. Among the regions composed of pixels at positions where the brightness difference is smaller than the predetermined brightness difference, the stain candidate image data representing the stain candidate contour image in which the pixel at the outermost edge of the region is a candidate for the pixel representing stain, that is, Dirt candidate contour image data is generated. Since the back surface of the polycrystalline silicon wafer 2 is imaged after the polycrystalline silicon wafer 2 is inverted 180 degrees by the polycrystalline silicon wafer reversing device 30, the back surface of the polycrystalline silicon wafer 2 is vertically Inverted. Therefore, in order to match the position of each pixel of the image indicated by the dirt candidate contour image data on the back surface with the position of each pixel of the image indicated by the dirt candidate contour image data on the front surface, the computer 10 The data is converted into inverted contour image data representing the dirt candidate contour image that is vertically inverted.

  FIG. 9 shows an example of a dirt candidate contour image on the surface of the polycrystalline silicon wafer 2. FIG. 9 shows a dirt candidate contour image generated based on the low-angle image of the surface and the high-angle image of the surface. In this embodiment, since the dirt is attached to the surface, The outline of the dirt portion is shown as the outline of the dirt candidate. In FIG. 9, the outline 73 of the dirt candidate due to the dirt is shown in the region 63.

  FIG. 10 shows an example of a reverse contour image of the back surface of the polycrystalline silicon wafer 2. FIG. 10 shows the reverse contour image of the back surface indicated by the reverse contour image data. In this other embodiment, no dirt is attached to the back surface, so that only the contour of the image of the grain boundary portion is a candidate for the stain. Shown as a contour. The inverted contour image data represents a soil candidate contour image obtained by vertically inverting the soil candidate contour image data generated based on the image data of the low-angle image and the image data of the high-angle image on the back surface of the polycrystalline silicon wafer 2. It is converted into image data. In the region 64 shown in FIG.

  The computer 10 compares the stain candidate contour image data on the front surface and the reverse contour image data on the back surface for each pixel at a position where the two surfaces face each other, and data indicating whether or not the pixel candidate represents a stain. Inconsistent image data representing an image in which the pixel at the position of the polycrystalline silicon wafer 2 corresponding to the position of the opposing pixel that is inconsistent is used as a mismatched pixel is generated. Then, it is determined that there is dirt when the contour of the region made up of mismatched pixels among the pixels of the image indicated by the mismatched image data satisfies a predetermined dirt condition.

  The computer 10 corresponds to the contour of the dirt candidate included in the candidate contour image on the back surface, or the contour satisfying the predetermined soil condition is a contour corresponding to the contour of the dirt candidate included in the front surface candidate contour image. Determine if it is a contour. If the contour is a contour corresponding to the contour of the stain candidate in the surface stain candidate contour image, it is specified that the stain is attached to the surface of the polycrystalline silicon wafer 2, or the contour is a stain candidate on the back surface. If the contour corresponds to the contour of the contour candidate of the contour image, it can be specified that the surface of the polycrystalline silicon wafer 2 is soiled.

  FIG. 11 shows an image obtained by enlarging the region 63 in the dirt candidate contour image shown in FIG. A region 63 is a part of the contamination candidate contour image on the surface of the polycrystalline silicon wafer 2. The area 63 is an area composed of 8 pixels in the vertical and horizontal directions. One cell indicates one pixel, a black pixel is a stain candidate pixel, and a white pixel is a pixel that is not a stain candidate. In a region 63, a contour 73 of a contamination candidate due to contamination is shown. The outline of the dirt candidate is the outline of the dirt or grain boundary.

  FIG. 12 shows an image obtained by enlarging the region 64 in the dirt candidate contour image shown in FIG. The region 64 is a part of the reverse contour image of the back surface of the polycrystalline silicon wafer 2 and corresponds to the region 63 shown in FIG. Similar to the region 63, the region 64 is a region composed of eight pixels in the vertical and horizontal directions. One cell indicates one pixel, a black pixel is a stain candidate pixel, and a white pixel is a pixel that is not a stain candidate. . The region 64 does not include the outline of the dirt candidate.

  FIG. 13 shows an image of mismatched image data obtained by comparing the region 63 shown in FIG. 11 with the region 64 shown in FIG. The mismatched image data is the pixel at the same position, that is, the position where the two surfaces face each other, where the computer 10 has the dirt candidate contour image data on the front surface and the inverted contour image data obtained by vertically inverting the dirt candidate contour image data on the back surface. This is image data representing a dirt candidate contour image in which a pixel at a position corresponding to the position of an opposing pixel where the data indicating whether or not it is a pixel candidate representing dirt is a mismatch is compared with each other. . In other words, the pixel-by-pixel data of the front surface dirt candidate contour image data and the pixel-by-pixel data of the reverse surface reverse contour image data are data obtained by exclusive ORing the pixels at the same position. FIG. 13 shows an image of the region 63 shown in FIG. 11, that is, a part of the surface candidate stain contour image shown in FIG. 9, and an image of the region 64 shown in FIG. 12, ie, the reverse of the back surface shown in FIG. This is an image obtained by taking an exclusive OR of corresponding parts of the contour image for each pixel at the same position.

  The predetermined dirt condition for judging that there is dirt is that the contour of the region composed of non-matching pixels is a contour in which eight pixels or more are adjacent to each other in, for example, an area of 8 pixels in the vertical direction and 8 pixels in the horizontal direction, and This is a condition that the pixels in the direction and the pixels in the horizontal direction are contours adjacent to each other by 3 pixels or more.

  The contour 83 of the region composed of non-matching pixels shown in FIG. 13 is a contour in which 18 pixels are adjacent to each other within the area of 8 pixels in the vertical direction and 8 pixels in the horizontal direction. Each contour is adjacent to three or more pixels. In the computer 10, the contour 83 of the region composed of non-matching pixels shown in FIG. 13 is a region composed of 16 or more pixels within a predetermined dirt condition, that is, within an area of 8 pixels vertically and 8 pixels horizontally. In addition, since the condition that the pixels in the vertical direction and the pixels in the horizontal direction are regions adjacent to each other by 3 pixels or more is satisfied, it is determined that the polycrystalline silicon wafer 2 is contaminated.

  In the computer 10, the contour 83 of the region composed of non-matching pixels shown in FIG. 13 corresponds to the contour 73 of the contamination candidate due to the contamination shown in FIG. 11, so that the contamination adheres to the surface of the polycrystalline silicon wafer 2. Is identified.

  In this way, light is irradiated by sequentially irradiating light from two angles to each surface, for example, a low angle of 10 to 30 degrees and a high angle of 60 to 80 degrees to the front or back surface. Images of each surface are picked up at each angle, and image data picked up from two angles, for example, low-angle image and high-angle image data, are converted into pixels at the same position on a polycrystalline semiconductor wafer, such as a polycrystalline silicon wafer 2. A brightness difference is calculated every time, and among the areas composed of pixels at a position where the calculated brightness difference is a brightness difference smaller than a predetermined brightness difference, a pixel at the outermost edge of the area is set as a pixel candidate representing dirt. Since the dirt candidate image data representing the dirt candidate contour image to be generated is generated for each surface, the dirt candidate image data obtained by extracting the contour of the portion where the brightness difference of the image data is small can be generated. Therefore, it is possible to extract the contours of grain boundaries and dirt portions having a small brightness difference. Since it is determined whether or not a predetermined dirt condition is satisfied by the contour, the determination can be made with a smaller amount of information than when determining by the region.

  FIG. 14 is a flowchart showing processing steps by an appearance inspection method according to another embodiment of the present invention. This flowchart shows a process to be processed by the polycrystalline silicon wafer visual inspection apparatus 1. When the polycrystalline silicon wafer 2 is placed on the stage of the transfer device 50, the process proceeds to step S1.

  In step S1, the angle of light irradiated from the light source unit 21 to the surface of the polycrystalline silicon wafer 2 is changed, two types of images of the surface are taken, and further, the back surface of the polycrystalline silicon wafer 2 is irradiated from the light source unit 41. Two kinds of images of the back surface are taken by changing the angle of the light to be emitted. The two types of images are a low angle image captured by irradiating light at an angle of 10 degrees to 30 degrees with respect to the front or back surface, and a high angle captured by irradiating light at an angle of 60 degrees to 80 degrees. There are two types of images. The captured images are stored in the hard disk as a low-angle image and a high-angle image on the front surface, and a low-angle image and a high-angle image on the back surface.

  In step S2, a low-angle image and a high-angle image of the surface stored in the hard disk are calculated for each pixel at the same position on the polycrystalline silicon wafer 2, and the calculated brightness difference is determined based on a predetermined brightness difference. Dirt candidate image data representing a dirt candidate image having a pixel at a small position as a candidate for a pixel representing dirt is generated, and the low-angle image and the high-angle image of the back surface stored in the hard disk are converted into the polycrystalline silicon wafer 2. A brightness difference is calculated for each pixel at the same position, and stain candidate image data representing a stain candidate image is generated with a pixel at a position where the calculated brightness difference is smaller than a predetermined brightness difference as a candidate for a pixel representing dirt. .

  In step S3, the stain candidate image data on the front surface and the stain candidate image data on the back surface are compared for each pixel at a position where the two surfaces face each other, and the data indicating whether or not the pixel candidate represents a stain is inconsistent. Inconsistent image data representing an image in which the pixel at the position of the polycrystalline silicon wafer 2 corresponding to the position of the opposing pixel is a mismatched pixel is generated.

  In step S <b> 4, it is determined whether or not a region composed of mismatched pixels among the pixels of the image indicated by the mismatched image data satisfies a predetermined dirt condition. When the predetermined dirt condition is satisfied, the process proceeds to step S5, and when the predetermined dirt condition is not satisfied, the process proceeds to step S6. In step S5, it is determined that there is dirt and the process ends. In step S6, it is determined that there is no dirt and the process ends.

  As described above, since the respective image data based on the images obtained by imaging the two surfaces of the polycrystalline semiconductor wafer such as the polycrystalline silicon wafer, that is, the front surface and the back surface are compared, the images appearing at the positions where both surfaces are opposed to each other are canceled out. Only images that appear on any one of the surfaces can be extracted. Therefore, even in a polycrystalline semiconductor wafer having a grain boundary, it is possible to cancel the pattern due to the grain boundary appearing on both surfaces, and to detect the dirt adhering to any one surface.

  Further, this appearance inspection method can detect dirt adhered to any one of the two surfaces, that is, the front surface or the back surface. The appearance inspection of a semiconductor wafer such as a polycrystalline silicon wafer 2 can be automated. Further, since it is possible to determine the presence / absence of dirt even on a thin polycrystalline semiconductor wafer that cannot be handled manually, a wafer without dirt can be provided to the cell manufacturing process of the solar cell.

  Furthermore, by comparing the candidate stain image data of two surfaces, that is, the candidate stain image data on the front surface and the candidate stain image data on the back surface, for each pixel at a position where the two surfaces face each other, Non-matching image data is generated by using a pixel at the position of a polycrystalline semiconductor wafer corresponding to the position of the opposing pixel where the data indicating whether or not there is a mismatch as the non-matching pixel. When the region satisfies the predetermined contamination condition, it is determined that there is contamination, so even if the grain boundary portion of the polycrystalline semiconductor wafer, for example, the polycrystalline silicon wafer 2, appears in the image indicated by the two surface contamination candidate image data, The grain boundary part is canceled out and does not become a mismatched pixel, but only the dirty part can be extracted as a mismatched pixel. When Mel dirty condition is satisfied, it can be determined that there is contamination. Therefore, even in a polycrystalline semiconductor wafer having a grain boundary, it is possible to cancel the pattern due to the grain boundary appearing on both surfaces, and to detect the dirt adhering to any one surface.

  Furthermore, each surface is sequentially irradiated with light from two angles, an image of each surface is captured for each angle at which the light is irradiated, and image data captured from each of the two angles, such as a low-angle image and A pixel at a position where the brightness difference is calculated for each pixel at the same position of the polycrystalline semiconductor wafer, for example, the polycrystalline silicon wafer 2, from the image data of the high-angle image, and the calculated brightness difference is smaller than a predetermined brightness difference. Is generated for each surface, and candidate stain image data obtained by extracting a portion having a small lightness difference in the image data can be generated. Therefore, it is possible to extract only the grain boundary and the soiled portion where the brightness difference is small.

  FIG. 15 is a flowchart showing processing steps by an appearance inspection method which is still another embodiment of the present invention. This flowchart shows a process to be processed by the polycrystalline silicon wafer visual inspection apparatus 1. When the polycrystalline silicon wafer 2 is placed on the stage of the transfer device 50, the process proceeds to step T1.

  In step T1, the angle of light irradiated from the light source unit 21 to the surface of the polycrystalline silicon wafer 2 is changed, two types of images of the surface are taken, and further, the back surface of the polycrystalline silicon wafer 2 is irradiated from the light source unit 41. Two kinds of images of the back surface are taken by changing the angle of the light to be emitted. The two types of images are a low angle image captured by irradiating light at an angle of 10 degrees to 30 degrees with respect to the front or back surface, and a high angle captured by irradiating light at an angle of 60 degrees to 80 degrees. There are two types of images. The captured images are stored in the hard disk as a low-angle image and a high-angle image on the front surface, and a low-angle image and a high-angle image on the back surface.

  In step T2, a lightness difference is calculated for each pixel at the same position on the polycrystalline silicon wafer 2 from the low-angle image and the high-angle image of the surface stored in the hard disk, and the calculated lightness difference is determined based on a predetermined lightness difference. Dirt candidate image data representing dirt candidate contour images, that is, dirt candidate contour image data, is generated by using, as a candidate for a pixel representing dirt, a pixel at the outermost edge of the area consisting of pixels at small positions. Further, a low-angle image and a high-angle image of the back surface stored in the hard disk are calculated for each pixel at the same position on the polycrystalline silicon wafer 2, and the calculated brightness difference is smaller than a predetermined brightness difference. Dirt candidate image data representing dirt candidate contour images, that is, dirt candidate contour image data, is generated by using, as a pixel candidate representing dirt, a pixel at the outermost edge of the area of pixels.

  In step T3, the front surface dirt candidate contour image data and the back surface dirt candidate contour image data are compared for each pixel at a position where the two surfaces face each other, and data indicating whether or not the pixel candidate represents a dirt. Disagreement image data representing an image in which the pixel at the position of the polycrystalline silicon wafer 2 corresponding to the position of the opposing pixel having a mismatch is used as the mismatch pixel is generated.

  In step T4, it is determined whether or not the contour of the region made up of mismatched pixels among the pixels of the image indicated by the mismatched image data satisfies a predetermined dirt condition. When the predetermined dirt condition is satisfied, the process proceeds to step T5, and when the predetermined dirt condition is not satisfied, the process proceeds to step T6. In step T5, it is determined that there is dirt and the process ends. In step T6, it is determined that there is no dirt and the process ends.

  In this way, each surface is sequentially irradiated with light from two angles, an image of each surface is captured for each angle at which light is irradiated, and image data captured from each of the two angles, for example, a low angle The image data of the image and the high-angle image are calculated for each pixel at the same position on the polycrystalline semiconductor wafer, for example, the polycrystalline silicon wafer 2, and the calculated brightness difference is a position where the calculated brightness difference is smaller than the predetermined brightness difference. Among the areas consisting of the pixels, the dirt candidate image data representing the dirt candidate contour image, which uses the pixel at the outermost edge of the area as the candidate for the pixel representing the dirt, is generated for each surface. It is possible to generate dirt candidate image data in which the outline of a small part is extracted. Therefore, it is possible to extract the contours of grain boundaries and dirt portions having a small brightness difference. Since it is determined whether or not a predetermined dirt condition is satisfied by the contour, the determination can be made with a smaller amount of information than when determining by the region.

  In any of the above-described embodiments, the light source unit 21 and the light source unit 41 irradiate light simultaneously from the light sources included in the four parts installed so as to surround the polycrystalline silicon wafer 2 from four sides, that is, in four directions. However, it is not always necessary to irradiate light from four directions at the same time, and light may be emitted from each direction sequentially. Although the number of images to be captured increases and the amount of image data transferred to the computer 10 increases, the amount of brightness difference information calculated for each pixel at the same position increases, so that the accuracy of the brightness difference is improved. Can improve the accuracy of detecting dirt.

  Further, in any of the above-described embodiments, each portion of the light source unit 21 and the light source unit 41 emits light at an angle of 10 degrees to 30 degrees with respect to two light sources, that is, the front surface or the back surface of the polycrystalline silicon wafer 2. Two types of images are emitted by irradiating light, ie, a low-angle image and a high-angle image, by a light source that irradiates light and a light source that irradiates light at an angle of 60 to 80 degrees with respect to the front or back surface of the polycrystalline silicon wafer However, the polycrystalline silicon wafer 2 can be irradiated by increasing the light source so that the light can be irradiated at an angle other than 10 ° to 30 ° and 60 ° to 80 °. You may increase the kind of image which images the front surface and back surface. Although the number of images to be captured increases and the amount of image data transferred to the computer 10 increases, the amount of brightness difference information calculated for each pixel at the same position increases, so that the accuracy of the brightness difference is improved. Can improve the accuracy of detecting dirt.

1 schematically shows a configuration of a polycrystalline silicon wafer visual inspection apparatus 1 according to an embodiment of the present invention. An example of the low angle image of the surface of the polycrystalline silicon wafer 2 imaged by the surface imaging device 20 shown in FIG. 1 is shown. An example of the high-angle image of the surface of the polycrystalline silicon wafer 2 imaged by the surface imaging device 20 shown in FIG. 1 is shown. An example of the dirt candidate image on the surface of the polycrystalline silicon wafer 2 is shown. An example of an inverted image obtained by vertically inverting the dirt candidate image on the back surface of the polycrystalline silicon wafer 2 is shown. The image which expanded the area | region 61 in the stain | pollution | contamination candidate image shown in FIG. 4 is shown. The image which expanded the area | region 62 in the stain | pollution | contamination candidate image shown in FIG. 5 is shown. 7 shows an image of mismatched image data obtained by comparing the region 61 shown in FIG. 6 with the region 62 shown in FIG. An example of the dirt candidate outline image of the surface of the polycrystalline silicon wafer 2 is shown. An example of the reverse outline image of the back surface of the polycrystalline silicon wafer 2 is shown. 10 shows an image obtained by enlarging a region 63 in the dirt candidate contour image shown in FIG. 9. The image which expanded the area | region 64 in the stain | pollution | contamination candidate outline image shown in FIG. 10 is shown. 11 shows an image of mismatched image data obtained by comparing the region 63 shown in FIG. 11 and the region 64 shown in FIG. It is a flowchart which shows the process process by the external appearance inspection method which is another form of implementation of this invention. It is a flowchart which shows the process process by the external appearance inspection method which is other form of implementation of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Polycrystalline silicon wafer visual inspection apparatus 2 Polycrystalline silicon wafer 10 Computer 20 Surface imaging device 21, 41 Light source part 22, 42 Camera 23, 43 Diffusion plate 24, 44 Bridge pier 30 Polycrystalline silicon wafer inversion device 31 Air suction device 32 Rotation Bar 40 Back side imaging device 50 Conveying device 71 Dirt candidate region 73 Dirt candidate outline

Claims (8)

An imaging step of individually imaging two opposing surfaces of the polycrystalline semiconductor wafer;
The respective image data based on the images of the two surfaces captured in the imaging process are compared for each pixel at a position where the two surfaces are opposed to each other, and mismatch image data representing an image having a mismatch pixel as a mismatch pixel is obtained. A comparison process to generate ;
Judgment to determine that one of the two surfaces is soiled when a region of the unmatched pixels among the pixels of the image indicated by the unmatched image data generated in the comparison step satisfies a predetermined stain condition and a step seen including,
The predetermined dirt condition is a second region in which a region composed of non-matching pixels is a region composed of pixels equal to or more than a first predetermined number of pixels, and a vertical pixel and a horizontal pixel are respectively predetermined. A method for inspecting the appearance of a polycrystalline semiconductor wafer, characterized in that it is a region adjacent to the number of pixels or more . The image data is dirt candidate image data indicating, for each pixel, whether or not the pixel candidate represents dirt.
In the comparison step, the stain candidate image data of the two surfaces is compared for each pixel at a position where the two surfaces are opposed to each other, and data indicating whether or not the pixel candidate represents a stain is opposed to each other. polycrystalline pixel position of the semiconductor wafer, the appearance inspection method of the polycrystalline semiconductor wafer according to claim 1, wherein the mismatch image data and said Rukoto forming raw and mismatch pixel corresponding to the position of the pixel. The method further includes a generation step of generating the stain candidate image data,
In the imaging step, each surface is sequentially irradiated with light from a plurality of angles, and images of each surface are captured for each angle at which the light is irradiated.
In the generation step, the brightness difference is calculated for each pixel at the same position of the polycrystalline semiconductor wafer, and the calculated brightness difference is smaller than the predetermined brightness difference. 3. The method for inspecting the appearance of a polycrystalline semiconductor wafer according to claim 2, wherein image data of a dirt candidate image having a pixel at a position that is a difference as the pixel candidate is generated as the dirt candidate image data for each surface. . The method further includes a generation step of generating the stain candidate image data,
In the imaging step, each surface is sequentially irradiated with light from a plurality of angles, and images of each surface are captured for each angle at which the light is irradiated.
In the generation step, the brightness difference is calculated for each pixel at the same position of the polycrystalline semiconductor wafer, and the calculated brightness difference is smaller than the predetermined brightness difference. Image data of a dirt candidate contour image having a pixel at the outermost edge of the area as a pixel candidate out of areas composed of pixels at positions that are differences is generated as the dirt candidate image data for each surface. A method for inspecting the appearance of a polycrystalline semiconductor wafer according to claim 2. Imaging means for individually imaging two opposing surfaces of the polycrystalline semiconductor wafer;
The respective image data based on the images of the two surfaces captured by the imaging means are compared for each pixel at a position where the two surfaces are opposed to each other, and mismatch image data representing an image having a mismatch pixel as a mismatch pixel is obtained. A comparison means to generate ;
Judgment to determine that one of the two surfaces is soiled when a region including the unmatched pixels satisfies a predetermined stain condition among the pixels of the image indicated by the mismatch image data generated by the comparison unit and it means only including,
The predetermined dirt condition is a second region in which a region composed of non-matching pixels is a region composed of pixels equal to or more than a first predetermined number of pixels, and a vertical pixel and a horizontal pixel are respectively predetermined. An appearance inspection apparatus for a polycrystalline semiconductor wafer, characterized in that it is an area adjacent to at least the number of pixels . The image data is dirt candidate image data indicating, for each pixel, whether or not the pixel candidate represents dirt.
The comparison means compares the stain candidate image data of the two surfaces for each pixel at a position where the two surfaces are opposed to each other, and the data indicating whether or not the pixel candidates indicate the stain are opposed to each other. the pixel position of the polycrystalline semiconductor wafer corresponding to the position of the pixel, the appearance inspection apparatus of the polycrystalline semiconductor wafer according to claim 5, characterized in Rukoto the mismatch image data constituting raw and mismatched pixels. Generating means for generating the stain candidate image data;
Irradiation means for irradiating light from a plurality of angles to the two surfaces,
The imaging means captures an image of each surface for each angle at which light is irradiated by the irradiation means,
The generation unit calculates a brightness difference for each pixel at the same position on the polycrystalline semiconductor wafer, and the calculated brightness difference is smaller than a predetermined brightness difference. 7. The appearance inspection apparatus for a polycrystalline semiconductor wafer according to claim 6, wherein image data of a dirt candidate image having a pixel at a position that is a difference as the pixel candidate is generated as the dirt candidate image data for each surface. . Generating means for generating the stain candidate image data;
Irradiation means for irradiating light from a plurality of angles to the two surfaces,
The imaging means captures an image of each surface for each angle at which light is irradiated by the irradiation means,
The generation unit calculates a brightness difference for each pixel at the same position on the polycrystalline semiconductor wafer, and the calculated brightness difference is smaller than a predetermined brightness difference. Image data of a dirt candidate contour image having a pixel at the outermost edge of the area as a pixel candidate out of areas composed of pixels at positions that are differences is generated as the dirt candidate image data for each surface. An appearance inspection apparatus for a polycrystalline semiconductor wafer according to claim 6.