JP6595800B2 - Defect inspection apparatus and defect inspection method - Google Patents

Description

  The present invention relates to a defect inspection apparatus and a defect inspection method, and more particularly to a defect inspection apparatus and a defect inspection method capable of detecting a defect in a cylindrical inspection object.

  When a machine tool such as a machine tool that processes an object to be processed with a processing tool performs processing such as cutting on the object to be processed, a defect such as wear or chipping may occur in the processing tool. When a defect occurs in the processing tool, it has an adverse effect such as deterioration of the processing accuracy of the processing object. In order to avoid this, it is necessary to inspect whether or not a defect has occurred and to use a machining tool that does not have a defect.

Patent Document 1 discloses a technique for inspecting whether or not a defect has occurred in a processing tool.
The tool defect inspection apparatus disclosed in Patent Document 1 includes a camera and an image processing apparatus. The camera captures a throw-away tip (work) such as a drill blade or a tap as a two-dimensional multivalued image (inspection image). The image processing apparatus compares a reference image (reference image) created in advance using a normal workpiece with an inspection image captured by the camera, extracts a defective portion, and determines the presence or absence of a defect.

  First, the image processing apparatus stores a reference image in advance for comparison with an image of a workpiece to be inspected. And the inspection image which consists of the outline and surface image of a workpiece | work is acquired from the camera which imaged the workpiece | work. Then, with respect to the inspection image of the work to be inspected, the brightness of the reference image prepared in advance for the collation is negative-positive inverted, the inspection image is overlapped with the reference image, a difference is obtained, and an average is obtained for each pixel. By this processing, all pixel data of the lightness value common to the reference image and the inspection image are canceled and become an intermediate density. A pixel value having an abnormal brightness, which is a brightness value different between the reference image and the inspection image, remains as a difference component, and this difference component can be clearly seen by the contrast with the intermediate density portion. Then, the difference image based on the difference component is binarized to obtain an extracted binarized image. When the area or dimension of the portion (defect candidate) indicating the abnormal pixel value in the extracted binarized image exceeds the specified range value, it is determined that there is a defect. Thereby, linear defects such as planar defects and scratches are also detected.

  As described above, the technique disclosed in Patent Document 1 captures a two-dimensional image with a camera and determines the presence or absence of a defect based on the image. For this reason, when the technique disclosed in Patent Document 1 is applied to a cylindrical workpiece having a processing blade or the like on the side surface, in order to detect a defect of the workpiece from the two-dimensional image, the imaging direction of the camera is changed to the cylindrical workpiece. It is necessary to fit the side of the. At this time, since the image is usually focused on the closest part of the camera on the side surface of the cylindrical workpiece, the focus is blurred except for the closest part of the camera. For this reason, in order to focus and image the entire workpiece side surface, it is necessary to perform a process for one round of the workpiece such as stopping the cylindrical workpiece and focusing and imaging a part of the side surface. For this reason, there exists a problem that measurement time becomes long in order to image the whole workpiece | work side surface.

  Patent Document 2 discloses a technique for inspecting whether or not a defect is generated in a cylindrical inspection object that can avoid this problem. The inspection device for defective appearance of a screw disclosed in Patent Document 2 acquires a one-dimensional light-dark figure by imaging a side surface of a screw along an axial direction with a line sensor while rotating the screw having a columnar shape. Then, the side surface of the screw is imaged over the entire area (for one round) and synthesized to create a binarized two-dimensional side surface developed image. Then, the side development image is compared with the side development image of a normal screw to determine whether or not a defect has occurred.

  As described above, the technique disclosed in Patent Document 2 uses a line sensor to continuously image while rotating a screw to create a binarized two-dimensional side developed image, and based on this, Detect defects. For this reason, the defect of a screw (inspection object) can be detected without increasing the measurement time.

JP 2007-278915 A JP 61-076941 A

  The tool defect inspection apparatus disclosed in Patent Document 1 described above uses a two-dimensional multi-value image as an inspection image, and takes the difference between the inspection image and the reference image as the multi-value image. Then, binarization is performed to create an extracted binarized image, and it is determined that there is a defect when the extracted binarized image has an area or size that is equal to or greater than a specified range value.

  For this reason, for example, there is a problem that even if there are a large number of quasi-defects having an area or size within a specified range value close to the specified range value in the extracted binary image, it is not determined that there is a defect. In this case, when this workpiece is used, there is a possibility that the machining accuracy may be adversely affected on the workpiece.

  In addition, the inspection device for a defective appearance of a screw disclosed in Patent Document 2 described above is a two-dimensional binarized image obtained by continuously imaging a side surface of a cylindrical screw while rotating the screw using a line sensor. A side development image is created, and a screw defect is detected based on this image. For this reason, it is possible to detect a defect of a cylindrical workpiece (screw) without increasing the measurement time.

  However, since this apparatus detects the occurrence of a defect from a binarized side surface developed image, there is a problem in that a minute defect below a binarization threshold cannot be detected.

  The object of the present invention is to solve the above-mentioned problems, and can detect minute defects (defects) on a cylindrical workpiece without increasing the measurement time, and it is determined that there are no defects even if there are many quasi-defects. It is to provide a defect inspection apparatus and a defect inspection method that do not cause a risk of failure.

  The defect inspection apparatus of the present invention captures a one-dimensional image of a side surface along the direction of the rotation axis while rotating a cylindrical inspection object illuminated from a predetermined direction about the rotation axis as a multi-value image. An output imaging unit and a side development image (inspection image) showing a composite image of the one-dimensional image for one round of the inspection target imaged by the imaging unit is created as a multi-valued image, and the side development image And a control unit that detects a defect of the inspection object based on a correlation value with a reference image indicating a side development image corresponding to a normal inspection object created in advance.

  The defect inspection method of the present invention captures a one-dimensional image of a side surface along the direction of the rotation axis while rotating a cylindrical inspection object illuminated from a predetermined direction about the rotation axis as a multi-value image. A side development image (inspection image) showing a composite image of the one-dimensional image for one round of the imaged inspection object is output and created as a multi-valued image, and the side development image and a normal inspection created in advance The defect of the inspection object is detected based on a correlation value with a reference image indicating a side development image corresponding to the object.

  ADVANTAGE OF THE INVENTION According to this invention, the defect inspection apparatus which can detect a micro defect (defect | deletion) with respect to a cylindrical workpiece | work without extending measurement time, and does not have a possibility of determining that there is no defect even if there are many quasi defects. And a defect inspection method can be provided.

It is a figure which shows an example of the defect inspection apparatus which concerns on the 1st Embodiment of this invention. It is a figure which shows an example of the defect inspection apparatus which concerns on the 2nd Embodiment of this invention. It is a flowchart which shows an example of operation | movement of the defect inspection apparatus which concerns on the 2nd Embodiment of this invention. It is an image figure which shows an example of a reference | standard image and a registration image pattern. It is an image figure which shows an example which produces the side surface expansion image before alignment from a one-dimensional brightness signal. It is an image figure which shows an example which shifts a part of side expansion | deployment image before alignment. It is a figure explaining an example of the coordinate of a reference image and a test | inspection image, a calculation range, etc. when calculating the correlation value of a reference | standard image and a test | inspection image.

Next, embodiments of the present invention will be described with reference to the drawings.
(First embodiment)
FIG. 1 is a diagram showing an example of a defect inspection apparatus according to the first embodiment of the present invention.

  The defect inspection apparatus 1 according to the present embodiment includes an imaging unit 2 and a control unit 3.

  The imaging unit 2 captures a one-dimensional image of a side surface along the rotation axis direction of a cylindrical inspection object illuminated from a predetermined direction, and outputs the image as a multi-valued image. The inspection object rotates about the rotation axis. The imaging unit 2 is, for example, a line sensor that repeatedly captures and outputs an image for one scan (one line). The control unit 3 creates a side development image (inspection image) showing a composite image of a one-dimensional image for one round of the inspection target imaged by the imaging unit 2 as a multi-value image. Then, based on a correlation value between the side developed image and a reference image indicating a side developed image corresponding to a normal inspection target created in advance, a defect of the inspection target is detected.

  First, a reference image indicating a side development image corresponding to the same thing as a normal inspection object without a defect is prepared in advance.

  Next, the cylindrical inspection object is illuminated from a predetermined direction (for example, a side surface). Then, the inspection object is rotated at a predetermined speed around the rotation axis.

  Then, the imaging unit 2 captures a one-dimensional image of the side surface along the direction of the rotation axis to be inspected and outputs it as a multi-valued image.

  The control unit 3 repeatedly receives the one-dimensional image of the inspection target imaged by the imaging unit 2, synthesizes these one-dimensional images, and generates a side developed image (inspection image) for one round of the inspection target as a multi-valued image. Create as is.

  Then, the control unit 3 calculates a correlation value between the created side developed image (inspection image) and a reference image prepared in advance. The correlation value is usually calculated including the pixel values around the pixel of interest. Then, based on the calculated correlation value, it is determined whether or not the inspection target has a defect.

  As described above, according to the first embodiment of the present invention, the imaging unit captures a one-dimensional image of the rotating cylindrical side surface of the inspection target and outputs it as a multi-valued image. Then, the control unit 3 creates a side development image (inspection image) to be inspected from the one-dimensional image output from the imaging unit as a multi-value image. Then, based on the correlation value between the inspection image and the reference image, a defect to be inspected is detected.

  For this reason, since the defect | deletion is detected with the multi-value image with respect to a cylindrical test object without binarizing, the micro defect | deletion below a binarization threshold value can also be detected.

  Further, the defect is detected based on the correlation value between the side-expanded image (inspection image) of the multi-valued image and the reference image. Since there is a difference between the correlation value when there are many quasi-defects and the correlation value when there are few (or no) quasi-defects, it is determined that there are no defects even if there are many quasi-defects. There is no fear.

Moreover, the multi-value image for detecting the defect | deletion of a test object is acquired by imaging the one-dimensional image of the side surface of the rotating cylindrical inspection object. For this reason, the measurement time of the multi-value image for detecting the defect | deletion of a test object does not become long.
(Second Embodiment)
FIG. 2 is a diagram showing an example of a defect inspection apparatus according to the second embodiment of the present invention.

  In the second embodiment of the present invention, a calculation formula for calculating the correlation value of the defect inspection apparatus 1 of the first embodiment is presented, and the first embodiment is more specific. FIG. 2 also shows a rotation control unit 4 that rotates the inspection object at a predetermined speed about the rotation axis. A description will be given mainly of the differences from the first embodiment.

  The defect inspection apparatus 1 according to the present embodiment includes an imaging unit 2 and a control unit 3.

  The correlation value is obtained by applying multi-value data of each pixel of the inspection image and the reference image to a predetermined calculation formula. The value of each pixel is, for example, a value (level) from 0 to 255 indicating luminance. The predetermined calculation formula is a general normal correlation calculation formula performed between multi-value data, or a correlation calculation formula modified based on this normal correlation calculation formula.

  The calculation formula is a formula for obtaining a normal correlation value at the coordinate position of the pixel for which the correlation between the inspection image and the reference image is obtained, and is a value obtained by dividing the covariance in the predetermined correlation calculation range at the coordinate position by the standard deviation. . This formula is a general normal correlation calculation formula performed between multi-value data. Further, a calculation formula (correlation calculation formula) obtained by modifying the standard deviation in the normal correlation calculation formula may be used.

  Next, operation | movement of the defect inspection apparatus 1 of this Embodiment is demonstrated using FIG.3, FIG.4, FIG.5, FIG.6 and FIG.

  FIG. 3 is a flowchart showing an example of the operation of the defect inspection apparatus according to the second embodiment of the present invention.

FIG. 4 is an image diagram illustrating an example of the reference image and the alignment image pattern.
The reference image is an image for calculating a correlation value with the inspection image corresponding to the processing tool to be inspected, and is configured by, for example, m × n pixel values. The white and gray parts are the parts of the machining tool. The white part indicates the cutting edge. Here, the luminance level of 30 or less is black (part A), 200 or more is white, and the others are gray. The alignment image pattern is an image pattern used when creating an inspection image corresponding to a processing tool, and is a part of a reference image. The position (α, β) and the size (range) of the image pattern are stored in advance in the storage unit together with the reference image.

  FIG. 5 is an image diagram showing an example of creating a side developed image (side developed image before alignment) from a one-dimensional light / dark signal. In this figure, the data (luminance level) of the part of the machining tool is shown in white and gray, the data of the part other than the machining tool is shown in black (part A), and the black data part is on the right side. It is trying to become. The one-dimensional light and dark signals of the machining tool repeatedly picked up by the image pickup unit 2 are arranged from the lower side in time series and developed into a two-dimensional image. A two-dimensional image having m × n pixels is created by arranging n light and dark signals of data length m.

  FIG. 6 is an image diagram showing an example of shifting a part of the side development image before alignment created in FIG. (A) is the side development image before the alignment produced in FIG. (B) is a side development image after alignment. An image pattern that matches the alignment image pattern shown in FIG. 4 is detected from the side development image before alignment created in FIG. 5, and the difference between the position of this image pattern and the position of the alignment image pattern (the amount of displacement) ) As a shift amount. Here, it is assumed that there is a positional shift only in the β direction. Then, an image corresponding to the shift amount is shifted with respect to the side development image before alignment to create a side development image (inspection image) after alignment. That is, the entire image is shifted downward so that the detected image pattern matches the alignment image pattern position. The lower part of the image after the shift is inserted in the upper part of the shortage after the shift.

FIG. 7 is a diagram for explaining an example of the coordinates and calculation ranges of the reference image and the inspection image when calculating the correlation value between the reference image and the inspection image.
(A) shows the coordinate (i ′, j ′) of the reference image, the coordinate position (α, β) of the pixel for which the correlation value with the inspection image is obtained, and the inspection image for the pixel at the coordinate position (α, β) The range (shaded area) for calculating the correlation value is shown. Yi′j ′ is the luminance level of the coordinates (i ′, j ′) of the pixel of the reference image, i ′ = i + α, j ′ = j + β, and (i, j) is The coordinates within the range for correlation calculation of the reference image. Here, i and j are 8, 16, 32, 64, etc., but may be appropriately determined according to the type and state of the workpiece, the lighting conditions, and the like.
(B) shows the coordinate (i ′, j ′) of the inspection image, the coordinate position (α, β) of the pixel for which the correlation value between the reference image and the reference image for the pixel at the coordinate position (α, β) The range (shaded area) for calculating the correlation value is shown. Xi′j ′ is the luminance level of the coordinate position (i ′, j ′) of the pixel of the inspection image, i ′ = i + α, j ′ = j + β, and (i, j) Is a coordinate within a range in which the correlation calculation of the inspection image is performed.

  A reference image corresponding to a processing tool to be inspected and an alignment image pattern used when creating an inspection image corresponding to the processing tool shown in FIG. 4 are stored in advance in a storage unit (not shown). . The reference image indicates a side development image corresponding to a normal processing tool having no defect among processing tools of the same type as the processing tool to be inspected. For example, the reference image may be created using the present apparatus (steps S1 to S4 in FIG. 3 described later). The alignment image pattern indicates a pattern of a characteristic part in the reference image, and is stored in the storage unit together with the position in the reference image.

  First, by the operator's operation, the processing tool to be inspected is set in the rotation control unit 4 so that the scanning line of the imaging unit 2 and the rotation axis of the processing tool are parallel to the front of the imaging unit 2. A holding portion other than the blade portion of the processing tool is held and fixed by a chuck or the like. And illumination is irradiated to a processing tool.

  In step S <b> 1 of FIG. 3, the control unit 3 causes the rotation control unit 4 to rotate the processing tool at a predetermined speed. This rotational speed is, for example, a speed at which the rotation angle during one scanning time when the imaging unit 2 captures one line is very small and does not cause out-of-focus, blurring, distortion, etc. in the captured image. It's okay.

  In step S2 of FIG. 3, the control unit 3 outputs an imaging start signal to the imaging unit 2, and the imaging unit 2 receives this signal and receives an image for one scan (one line) (one-dimensional multi-valued brightness / darkness). Signal) is repeated and imaged and output.

  In step S3 in FIG. 3, the control unit 3 causes the rotation control unit 4 to finish the rotation of the processing tool after a predetermined time has elapsed to complete the imaging of the entire circumference of the processing tool, and the imaging unit 2 performs imaging. Outputs a stop signal.

  In this way, since it is not necessary to stop the machining tool at a specific position during imaging, an image for one round of the machining tool can be taken at high speed.

  In step S4 of FIG. 3, as shown in FIG. 5, the control unit 3 arranges the one-dimensional light and dark signals repeatedly picked up by the image pickup unit 2 in chronological order, and shows a two-dimensional side surface representing one round of the processing tool. A developed image (side developed image before alignment) is created as a multi-valued image.

  In step S5 in FIG. 3, the control unit 3 shifts a part of the side development image before alignment and inspects it as shown in FIG. 6 in order to obtain a correlation value at the same position in the reference image. Create an image (side-expanded image after alignment) and align it with the reference image.

  In step S6 of FIG. 3, as shown in FIG. 7, the control unit 3 correlates the correlation value (degree of coincidence) between the reference image and the laterally developed image (inspection image) after alignment created in step S5. Calculate

  The calculation formula is a normal correlation calculation formula (Formula 1) for obtaining a normal correlation value at the coordinate position of the pixel for which the correlation between the inspection image and the reference image is obtained. Covariance in a predetermined correlation calculation range at the coordinate position is standard. The value divided by the deviation. This formula is a general normal correlation calculation formula performed between multi-value data.

Normal correlation value F1 (α, β) =
Σ ((Xi'j'-X) * (Yi'j'-Y)) / √ (Σ (Xi'j'-X) ² ) * √ (Σ (Yi'j'-Y) ² )
... (Formula 1)
Here, (α, β) is the coordinate position of the pixel for which the correlation between the inspection image and the reference image is obtained, and the normal correlation value F1 (α, β) is the normal correlation value for the pixel at the coordinate position (α, β). is there. X is the average level of pixels within the range of correlation calculation of the inspection image, Y is the average level of pixels within the range of correlation calculation of the reference image, and Xi′j ′ is the coordinate position (i ′ of the pixel of the inspection image) , J ′) level. Here, i ′ = i + α, j ′ = j + β, and (i, j) are coordinates within a range where the correlation calculation of the inspection image is performed. Yi′j ′ is the level of the coordinate position (i ′, j ′) of the pixel of the reference image, where i ′ = i + α, j ′ = j + β, (i, j) are the reference The coordinates within the range for correlation calculation of the image, and Σ is the total sum within the range for correlation calculation (the hatched portion in FIG. 7).

  Alternatively, the standard deviation indicating the denominator of Equation 1 may be modified so that the larger pixel value of the inspection image and the reference image is used.

  That is, a calculation formula obtained by modifying Formula 1 is a correlation calculation formula (Formula 2) for obtaining a correlation value at a coordinate position of a pixel for obtaining a correlation between the inspection image and the reference image. This calculation formula is obtained by dividing the covariance in the predetermined correlation calculation range at the coordinate position by the standard deviation modified to use the larger pixel value of the inspection image and the reference image for each correlation calculation range. It is.

Correlation value F2 (α, β) =
Σ ((Xi′j′−X) * (Yi′j′−Y)) / Σ (Max (Xi′j′−X, Yi′j′−Y)) ² (Equation 2)
Here, (α, β) is the coordinate position of the pixel for which the correlation between the inspection image and the reference image is obtained, and the correlation value F2 (α, β) is the correlation value for the pixel at the coordinate position (α, β). X is the average level of pixels within the range of correlation calculation of the inspection image, Y is the average level of pixels within the range of correlation calculation of the reference image, and Xi′j ′ is the coordinate position (i ′ of the pixel of the inspection image) , J ′) level. Here, i ′ = i + α, j ′ = j + β, and (i, j) are coordinates within a range where the correlation calculation of the inspection image is performed. Yi′j ′ is the level of the coordinate position (i ′, j ′) of the pixel of the reference image, where i ′ = i + α, j ′ = j + β, (i, j) are the reference This is the coordinates within the range for correlation calculation of images. Max () indicates an operation for calculating the higher level by comparing the parentheses (), and Σ is the total sum within the range for correlation calculation (the hatched portion in FIG. 7).

  In the case of using Expression 2, since the standard deviation using the larger pixel value of the inspection image and the reference image is used, in general, compared to Expression 1 for obtaining the normal correlation value, it is due to the deficiency. It is easily affected by deformation, that is, it is easy to detect a defect.

  The normal correlation value and the correlation value obtained by these formulas are values in the range of 1 to 0, and the closer to 1, the more similar the inspection image and the reference image, that is, the higher the matching degree. Yes.

  The normal correlation value or the correlation value is calculated for all coordinate positions (α, β) on the reference image and the inspection image, as shown in FIG.

  In step S7 of FIG. 3, the control unit 3 has coordinates (α, β) whose normal correlation value or correlation value (degree of coincidence) obtained in step S6 is a predetermined value (for example, 0.9) or less. In this case, it is assumed that there is a defect at the position of this coordinate, and this work is determined as a missing work. The predetermined value 0.9 is not limited to this value, and may be changed as appropriate by experiments or the like including the influence of the type and state of the workpiece, the lighting conditions, and the like.

  Thus, according to the second embodiment of the present invention, the following effect is obtained in addition to the effect obtained in the first embodiment. That is, two calculation formulas for obtaining the correlation between the inspection image and the reference image were prepared: a normal correlation calculation formula (Formula 1) and a correlation calculation formula (Formula 2). For this reason, the optimal expression can be selected from these expressions through experiments and the like, including the effects of the type and state of the workpiece, the lighting conditions, and the like. For this reason, it is possible to detect a minute defect more appropriately and effectively, and it is possible to determine that a defect having many quasi-defects is defective.

Part or all of the above-described embodiments can be described as in the following supplementary notes, but is not limited thereto.
(Appendix 1)
An imaging unit that captures a one-dimensional image of a side surface along the direction of the rotation axis while rotating a cylindrical inspection object illuminated from a predetermined direction around the rotation axis, and outputs the image as a multivalued image;
A side development image (inspection image) showing a composite image of the one-dimensional image for one round of the inspection object imaged by the imaging unit is created as a multi-value image, and the side development image and a normal examination created in advance A control unit for detecting a defect of the inspection object based on a correlation value with a reference image indicating a side development image corresponding to the object;
A defect inspection apparatus comprising:
(Appendix 2)
2. The defect inspection apparatus according to claim 1, wherein the correlation value is obtained by applying multi-value data of each pixel of the inspection image and the reference image to a predetermined calculation formula.
(Appendix 3)
The calculation formula is a formula obtained by dividing the covariance in the predetermined correlation calculation range at the coordinate position by the standard deviation.
The defect inspection apparatus according to Supplementary Note 2, wherein:
(Appendix 4)
The calculation formula is as follows: normal correlation value F1 (α, β) =
Σ ((Xi'j'-X) * (Yi'j'-Y)) / √ (Σ (Xi'j'-X) ² ) * √ (Σ (Yi'j'-Y) ² )
here,
(α, β) is the coordinate position of the pixel for which the correlation between the inspection image and the reference image is obtained,
The normal correlation value F1 (α, β) is a normal correlation value for the pixel at the coordinate position (α, β),
X is the average level of pixels within the range for correlation calculation of the inspection image,
Y is the average level of pixels within the range for correlation calculation of the reference image,
Xi′j ′ is the level of the coordinate position (i ′, j ′) of the pixel of the inspection image, where i ′ = i + α, j ′ = j + β, (i, j) is It is the coordinates within the range for correlation calculation of the inspection image,
Yi′j ′ is the level of the coordinate position (i ′, j ′) of the pixel of the reference image, where i ′ = i + α, j ′ = j + β, (i, j) is The coordinates within the range of correlation calculation of the reference image,
Σ is the sum within the range for correlation calculation,
The defect inspection apparatus according to Supplementary Note 3, wherein
(Appendix 5)
The calculation formula is a standard deviation obtained by transforming the covariance in the predetermined correlation calculation range at the coordinate position to use the larger pixel value of the inspection image and the reference image for each correlation calculation range. Divided by
The defect inspection apparatus according to supplementary note 3 or 4, characterized in that.
(Appendix 6)
The calculation formula is as follows: correlation value F2 (α, β) =
Σ ((Xi′j′−X) * (Yi′j′−Y)) / Σ (Max (Xi′j′−X, Yi′j′−Y)) ²
here,
(α, β) is the coordinate position of the pixel for which the correlation between the inspection image and the reference image is obtained,
The correlation value F2 (α, β) is a correlation value for the pixel at the coordinate position (α, β),
X is the average level of pixels within the range for correlation calculation of the inspection image,
Y is the average level of pixels within the range for correlation calculation of the reference image,
Xi′j ′ is the level of the coordinate position (i ′, j ′) of the pixel of the inspection image, where i ′ = i + α, j ′ = j + β, (i, j) is It is the coordinates within the range for correlation calculation of the inspection image,
Yi′j ′ is the level of the coordinate position (i ′, j ′) of the pixel of the reference image, where i ′ = i + α, j ′ = j + β, (i, j) is The coordinates within the range of correlation calculation of the reference image,
Max () indicates the operation to calculate the higher level by comparing ()
Σ is the sum within the range for correlation calculation,
The defect inspection apparatus according to appendix 5, characterized by:
(Appendix 7)
The control unit causes the imaging unit to capture a plurality of the one-dimensional images while rotating the cylindrical inspection object about the rotation axis, and creates the side-expanded image based on the one-dimensional images. To
The defect inspection apparatus according to appendix 1, 2, 3, 4, 5 or 6 characterized by the above.
(Appendix 8)
Taking a one-dimensional image of a side surface along the direction of the rotation axis while rotating a cylindrical inspection object illuminated from a predetermined direction around the rotation axis, and outputting as a multi-valued image,
A side development image (inspection image) showing a composite image of the one-dimensional image for one round of the imaged inspection object is created as a multi-value image,
Detecting a defect of the inspection object based on a correlation value between the side development image and a reference image indicating a side development image corresponding to a normal inspection object created in advance;
A defect inspection method characterized by that.
(Appendix 9)
9. The defect inspection method according to appendix 8, wherein the correlation value is obtained by applying multi-value data of each pixel of the inspection image and the reference image to a predetermined calculation formula.
(Appendix 10)
The calculation formula is a formula obtained by dividing the covariance in the predetermined correlation calculation range at the coordinate position by the standard deviation.
The defect inspection method according to appendix 9, wherein
(Appendix 11)
The calculation formula is a standard deviation obtained by transforming the covariance in the predetermined correlation calculation range at the coordinate position to use the larger pixel value of the inspection image and the reference image for each correlation calculation range. Divided by
The defect inspection method according to supplementary note 10, wherein:

DESCRIPTION OF SYMBOLS 1 Defect inspection apparatus 2 Imaging part 3 Control part 4 Rotation control part

Claims (4)

An imaging unit that captures a one-dimensional image of a side surface along the direction of the rotation axis while rotating a cylindrical inspection object illuminated from a predetermined direction around the rotation axis, and outputs the image as a multivalued image;
A side development image (inspection image) showing a composite image of the one-dimensional image for one round of the inspection object imaged by the imaging unit is created as a multi-value image, and the side development image and a normal examination created in advance A control unit for detecting a defect of the inspection object based on a correlation value obtained by applying multi-value data of each pixel of a reference image indicating a side development image corresponding to the object to a predetermined calculation formula ;
A defect inspection apparatus comprising:
The calculation formula is obtained by dividing the covariance in the predetermined correlation calculation range at the coordinate position of each pixel by the standard deviation, and the standard deviation is calculated from the inspection image and the reference image for each correlation calculation range. It is modified to use the larger pixel value of
A defect inspection apparatus characterized by that. The calculation formula is as follows: correlation value F2 (α, β) =
Σ ((Xi′j′−X) * (Yi′j′−Y)) / Σ (Max (Xi′j′−X, Yi′j′−Y)) ²
here,
(α, β) is the coordinate position of the pixel for which the correlation between the inspection image and the reference image is obtained,
The correlation value F2 (α, β) is a correlation value for the pixel at the coordinate position (α, β),
X is the average level of pixels within the range for correlation calculation of the inspection image,
Y is the average level of pixels within the range for correlation calculation of the reference image,
Xi′j ′ is the level of the coordinate position (i ′, j ′) of the pixel of the inspection image, where i ′ = i + α,
j ′ = j + β, (i, j) are the coordinates within the range of the correlation calculation of the inspection image,
Yi′j ′ is the level of the coordinate position (i ′, j ′) of the pixel of the reference image, where i ′ = i + α,
j ′ = j + β, (i, j) are coordinates within the range of correlation calculation of the reference image,
Max () indicates the operation to calculate the higher level by comparing ()
Σ is the sum within the range for correlation calculation,
The defect inspection apparatus according to claim 1. The control unit causes the imaging unit to capture a plurality of the one-dimensional images while rotating the cylindrical inspection object around the rotation axis, and creates the side-expanded image based on the one-dimensional images. To
The defect inspection apparatus according to claim 1 or 2, wherein Taking a one-dimensional image of a side surface along the direction of the rotation axis while rotating a cylindrical inspection object illuminated from a predetermined direction around the rotation axis, and outputting as a multi-valued image,
A side development image (inspection image) showing a composite image of the one-dimensional image for one round of the imaged inspection object is created as a multi-value image,
Based on the correlation value obtained by applying the multi-value data of each pixel of the reference image indicating the side development image and the side development image corresponding to the normal inspection object created in advance to a predetermined calculation formula, the defect of the inspection object A defect inspection method for detecting
The calculation formula is obtained by dividing the covariance in the predetermined correlation calculation range at the coordinate position of each pixel by the standard deviation, and the standard deviation is calculated from the inspection image and the reference image for each correlation calculation range. It is modified to use the larger pixel value of
A defect inspection method characterized by that .

Images