JP2006153775A5 - - Google Patents

Description

Misalignment detection method

The present invention relates to a positional deviation detecting how the transparent member when assembling the at least one transparent member.

Normally, when assembling two optical elements, it is necessary to align the optical elements. Patent Documents 1 and 2 disclose alignment between liquid crystal display cells and alignment between liquid crystal substrates. However, in these alignment methods, since the absolute position of the mark attached to the object (cell or substrate) positioned above and below is individually obtained and the amount of displacement is calculated therefrom, Therefore, there is a problem that the calculation processing becomes slow.
JP 2000-180810 A JP 2004-151215 A

An object of the present invention, provides a positional deviation detecting how that can be detected positional deviation amount the position of the object using the image processing techniques such as optical elements and / or the positional deviation direction fast and efficient There is to do.

To achieve the above object, the present invention is the amount of deviation from a predetermined position of the position reference mark on the basis of the shade of an image obtained by observing the position reference mark formed on the object as the feature points and / or positional deviation direction detection A method for detecting misalignment, wherein two position reference marks are set, and one position reference mark is set as a predetermined position to detect a position shift amount and / or a position shift direction of the other position reference mark. Features.

When the object on which the position reference mark is formed is optically observed, the characteristic points observed as an image or the surroundings of the characteristic points have different shades from other regions. By detecting this, it is possible to detect the positional deviation amount and / or the positional deviation direction viewed from the observation direction of the position reference mark with high accuracy.

  In the positional deviation detection method according to the present invention, it is preferable that the position reference mark is observed by an image photographing unit. By capturing the feature points as an image, extraction of the observed feature point region and calculation of the displacement amount and / or displacement direction can be easily processed by the arithmetic device.

  A region where the observed brightness is larger or smaller than a predetermined value can be determined as a feature point, or a region showing a color within a predetermined range in the observed image is determined as a feature point. Can do. Then, the predetermined value can be obtained from the brightness distribution in the image, or the predetermined value can be obtained from the vicinity value of the average value of the brightness in the image, or the predetermined value can be obtained from the image in the image. It can be obtained from the vicinity of the most brightness in the distribution of brightness.

Further, in the positional deviation detection method according to the present invention, you two sets said position reference marks. In this case, the displacement direction and / or displacement amount can be detected based on the size of the two observed feature points. The misalignment direction and / or misalignment amount may be detected based on the difference in brightness between the two observed feature points. Further, the position shift direction and / or the position shift amount may be detected based on the region including the two feature points and / or the shading around the region.

Alternatively, the position is based on an overlap image that appears when the first feature point overlaps part or all of the second feature point when viewed from the observation direction and / or a brightness deviation or gradient around the overlap image. The shift direction and / or the position shift amount can be detected.

Furthermore, the position shift direction and / or the position shift amount can be detected from the position difference between the two feature points, or the position shift direction and / or position shift based on the size of the region including the two feature points. The amount can be detected. In this case, the misalignment direction and / or misalignment amount can be detected based on the line segment connecting the centroids of the two feature points, and the region including the two feature points can be detected in the misalignment direction. The amount of misalignment can be detected based on the line segment formed by the cut section.

  Further, a region including two feature points may be a sum region of two feature point regions. The region including the two feature points may be an image region formed by overlapping the first feature point with the image of the second feature point. Further, the area including the two feature points may be a minimum rectangle including the two feature points.

In addition, it is possible to detect the shape of the two feature points and detect the displacement direction and / or the displacement amount. In this case, the center of each of the two feature point regions may be obtained using a detection method for detecting a circle, and the amount of displacement may be detected based on the distance between the centers. A Hough transform can be used as a detection method for recognizing and detecting a shape.

According to the positional deviation detecting how according to the present invention described above, by optically observing the position reference mark of the object as a feature point, the method of image processing can be accurately detected position, the object of the Position adjustment and tilt adjustment can be performed quickly and efficiently. In addition, it is possible to detect the displacement effectively without damaging the object, and it is possible to detect the displacement with an accuracy of about 10 to 50 times the resolution of the image.

Hereinafter, an embodiment of the displacement detecting how according to the present invention will be described with reference to the accompanying drawings.

(Refer 1st Example and FIGS. 1-15)
In FIG. 1, as the first embodiment, a displacement amount and / or a displacement direction of a position reference mark 111 formed on the surface of a transparent object to be adjusted 110 from a predetermined position is detected, and the object to be adjusted 110 is placed on the XY plane. The position adjustment device is shown in which the object to be adjusted 110 is centered with respect to the adjustment member 120 by being moved by. The adjustment object 110 is a transparent member that needs to be aligned, and the adjustment member 120 is a substrate for assembling the transparent member.

  In this position adjustment device, the position reference mark 111 of the object 110 to be adjusted placed on the adjustment member 120 and the position reference mark 121 formed on the adjustment member 120 are photographed by the observation device (CCD camera) 100 (FIG. In step S1), a feature point region is extracted by performing image processing on the captured feature point (step S2), and a positional deviation amount is calculated from the feature point region in the observation image or its surroundings (step S3). Then, the object to be adjusted 110 is moved on the XY plane with respect to the adjustment member 120 by the drive stage 107 based on the calculated value. The drive stage 107 is based on a well-known XY drive method.

  When the position reference marks 111 and 121 are optically observed, the feature points observed as an image or the surroundings of the feature points have different shading tendencies from other regions. By detecting this, it is possible to detect with high accuracy the amount of deviation seen from the observation direction of the mark 111 with respect to the mark 121.

  In the first embodiment, the position reference mark 111 has a convex shape with a diameter of 10 μm, and the position reference mark 121 is formed as a hole with a diameter of 10 μm. The result captured by the observation apparatus 100 having the optical axis P perpendicular to the XY direction is transferred to the CPU 105 as a 640 × 480 pixel 24-bit RGB color image and displayed on the display 106. The brightness L of the pixel can be expressed by the following equation using the respective color components (R, G, B).

  L = 0.298912 × R + 0.586611 × G + 0.114478 × B

  The feature points (photographed images of the position reference marks 111 and 121) photographed by the observation apparatus 100 are photographed such that the mark 111 of the object 110 to be adjusted is bright and the mark 121 of the adjustment member 120 is dark. The CPU 105 extracts a point that is higher than the value obtained by adding 10 to the average brightness value as the feature point region. The drive stage 107 is driven according to the result of processing the input image according to a predetermined parameter or the value designated by the operator.

  First, the drive stage 107 is manually operated to obtain the parameters a1 and a2 according to the procedure shown in FIG. That is, adjustment is performed so that the marks 111 and 121 (feature points) of the object to be adjusted 110 and the adjustment member 120 are substantially overlapped with each other (step S11), and the object to be adjusted 110 is moved by 5 μm in the X direction (step S12). Then, the minimum rectangle including the feature point region on the image is expanded by 5 pixels (step S13). Then, a deviation V of the brightness distribution is obtained within the square obtained in step S13 (step S14). Further, the object to be adjusted 110 is moved by 5 μm in the X direction (step S15), and the width x1 of the smallest rectangle including the feature point region on the image is measured (step S16). Further, the object to be adjusted 110 is moved in the X direction by 5 μm (step S17), and the minimum width x2 of the feature point area on the image is measured (step S18).

  The parameters a1 and a2 are expressed by the following equations.

a1 = 5 / Vx
a2 = 5 / | x1-x2 |
Vx is the X direction component in the image of deviation V

  The parameters a1 and a2 are set in the CPU 105, and the CPU 105 calculates the amount of displacement according to the parameters according to the procedure shown in FIG. In this calculation procedure, D is a deviation amount of the adjustment member 120 when the adjustment target 110 is used as a reference. dt is the required accuracy. Here, dt = 1 μm.

  First, the CPU 105 acquires an input image (step S21), and extracts a feature point area (step S22). Then, the smallest rectangle R that includes the feature point region is obtained (step S23), and it is determined whether the diagonal line of the rectangle R is smaller than x2 (step S24).

  If “YES” in the step S24, the square R is expanded by 10 pixels to obtain an ambient brightness deviation V (step S25), and a deviation amount D = a1V is obtained (step S26). On the other hand, if NO in step S24, the average values (centroids) Gb and Gd of the coordinates on the image of the points whose brightness is larger and smaller than the average value of the entire feature point region are obtained (step S27), and the deviation is detected. A quantity D = a2 (Gb−Gd) is obtained (step S28).

  Next, it is determined whether or not the deviation amount | D | obtained in step S26 or S28 is smaller than the required accuracy dt (step S29). If it is smaller, the operation is terminated. If larger, the drive stage 107 is given a value D as an instruction to drive (step S30), and the process returns to step S21.

  Note that the value of the required accuracy dt may exist individually in the X and Y directions. In this case, each element of the deviation amount D is compared with the threshold values in the X and Y directions. Here, the initial parameters are calculated in the X direction, but may be calculated in the Y direction. Moreover, although the same coefficient was used in the X and Y directions, the parameter 110 may be obtained individually in the X and Y directions by moving the object 110 to be adjusted in the Y direction. In addition, the initial parameter may be calculated only once for the first time as long as the displacement of the equivalent member is detected.

  In the first embodiment, since the position reference marks 111 and 121 are photographed by the observation apparatus 100 and handled as feature point images, the feature point region can be easily extracted by calculating the positional deviation amount. In addition, it is possible to effectively detect a positional shift without damaging the object to be adjusted 110 and the adjustment member 120, and it is possible to detect the shift amount with an accuracy of about 10 to 50 times the resolution of the image.

  Since the feature point region extracted as an image has a tendency of brightness different from other regions, the brightness is obtained from the shade of the image, and a place where the value is larger or smaller than a predetermined value is determined as the feature point region. Can do. The brightness index is the brightness index in the L * u * v color space, the pixel value itself in the case of a grayscale image, the arbitrary color component of the RGB / CMY value in the case of a color image, or the square of the RGB color component. The sum or the square root thereof, the brightness when converted into the HBS value, and the like can be arbitrarily selected.

  Also, depending on the material of the object on which the position reference mark is formed, the feature point region observed as an image tends to show a specific color. Since some errors may occur due to the influence of noise or the like, it is determined that similar colors are also feature point regions. For a predetermined color, a specific area of a predetermined color space is determined so that an area desired to be a feature point area is included in advance, and is selected and used. For example, when the feature point is blue and the color of other places indicates other than blue, the blue portion is detected as the feature point region. In addition, when the position reference mark is illuminated with a light beam mixed with light of different wavelengths, the mark image (feature point) can be observed as a region of a different color due to factors such as a difference in the refractive index of the mark or object. As the color, a value in a color space indicated by the CIE standard, an RGB / CMY value, an HSB value, or the like can be used.

  The predetermined value can be obtained from the distribution of brightness in the image. The brightness of the feature point area in the photographed image varies depending on the photographing condition of the position reference mark. Therefore, the feature point region can be effectively detected by extracting a relatively bright part or dark part using the brightness distribution of the entire image. Further, the predetermined value may be obtained from the vicinity value of the average brightness value in the image.

  Alternatively, the predetermined value can be obtained from the vicinity of the highest brightness in the brightness distribution in the image. The area occupied by the feature points in the captured image is less than half when viewed from the entire image. In addition, areas other than the feature points show almost the same brightness. Therefore, when the frequency distribution of brightness is taken, portions other than the feature point region appear with high frequency. That is, when the feature point region is extracted as a region that is relatively brighter or darker than other portions, the feature point region is a region that is brighter or darker than the average brightness value or maximum likelihood value of the entire image. Therefore, the brightness P of the pixels in the region extracted as the feature point region satisfies the following condition.

P ≧ Pt−Pg (when the feature point area is dark)
P ≦ Pt + Pg (when the feature point area is bright)

  Pg in the above formula is the gap Pg shown in FIG. 5, and Pt is the average value or maximum likelihood value of the entire image. Since the brightness of the feature point area has an average value or maximum likelihood value and some gap Pg, only the feature point area can be efficiently extracted by adjusting the gap Pg. The value of the gap Pg is preferably about 1 to 20% of the value of Pt.

  Further, in the first embodiment, since two position reference marks are set, the misalignment amount of the feature point is obtained by simultaneously photographing the marks 111 and 121, and the object to be adjusted 110, the adjustment member 120, Can be detected.

  It is also possible to obtain the posture and manufacturing accuracy of an object by forming marks at two points of one object and obtaining the relative deviation amount of the feature points observed as an image. In this case, the deviation amount of the feature point is calculated, and the result is converted into the actual deviation amount.

  In FIG. 6, (A) schematically illustrates a case where the amount of displacement between the two objects 150 and 160 is detected, and FIG. 6 (B) schematically illustrates a case where the amount of displacement between the front and back surfaces of the object 170 is detected. FIG. 6C schematically shows a case where the tilt amount of the object 180 is detected. Reference numerals 151, 161, 171, 172, 181, and 182 are position reference marks.

  Further, the misalignment direction can be detected by the magnitude of the two observed feature points. The feature point regions in the image of the two feature points appear to have different sizes depending on factors such as the position of the feature point, the observation position, and the focal length at the time of observation. By checking the size of both position reference marks in advance, it is possible to determine which mark each feature point area is based on, and thereby the position shift direction can be detected.

  Further, it is possible to detect the direction of displacement based on the difference in brightness between the two observed feature points. In the case of observation using illumination and an optical system, the observed two feature points show a difference in brightness depending on the focal length, the position of the mark, the material of the object, and the like. Therefore, by detecting the feature point area and comparing the average value and maximum likelihood of brightness within the area, it is possible to determine which mark each feature point area is based on. The misalignment direction can be detected.

  For example, as shown in FIG. 7A, when the position reference mark 171 formed on the translucent object 170 semi-transmits light and the position reference mark 172 absorbs light, both the marks 171 and 172 Although it is observed darker than the area, the mark 171 is observed brighter than the mark 172. Further, as shown in FIG. 7B, when the objects 150 and 160 are transparent, the position reference mark 151 is a protrusion formed on the object 150, and the position reference mark 161 is a hole formed on the object 160, the protrusion The portion (mark 151) causes the action of a lens, and the brightness of each feature point region in the image changes depending on the observation position.

  In FIGS. 7A and 7B, the illumination direction does not have to be opposite to the observation direction, and the illumination means is not necessarily required. For example, this is the case where the mark 151 shown in FIG. 7B emits light. Further, the mark 151 may reflect a light beam. In that case, an illumination unit may be used to enhance the reflected light if necessary.

  Further, it is possible to detect the position shift direction and / or the position shift amount based on the area including the two feature points and / or the density around the area. If the two feature points are very close or have a positional relationship such that some or all of them overlap, it is impossible to discriminate them based on the brightness and size of the feature points themselves. However, since characteristic shades appear in and around the region including the feature points, the displacement direction and the displacement amount can be detected based on the shades.

  Specifically, as shown in FIG. 8A, when two feature points partially overlap or are very close, as shown in FIG. 8B, when two feature points almost overlap, A characteristic shading distribution appears around it. By detecting the distribution, it is possible to determine the direction and distance of the relative displacement of the feature points.

  In addition, as shown in FIG. 9A, an overlapping image that appears when one feature point overlaps part or all of the other feature point as viewed from the photographing direction and / or brightness around the overlapping image. The misalignment direction and / or misalignment amount can be detected based on the deviation or gradient of the height. Such a positional deviation amount and a deviation in brightness distribution in and around the feature point region are in a substantially linear relationship. That is, when the brightness bias is expressed as a deviation or a gradient (see FIG. 9B), the deviation amount can be obtained by multiplying this by a constant. This constant can be obtained from the result of measuring the feature point several times at different positions. Since the area around the feature point area depends on the size of the feature point area that appears when they overlap, the size of the feature point area in the state where the feature points are almost completely overlapped in advance is checked. It is preferable to set several times as the surrounding area. The shape is not limited to a circle, and an arbitrary shape such as a quadrangle can be selected.

  Further, it is possible to detect the amount of positional deviation from the positional difference between the two feature points. In other words, the distance between the two feature points has a linear relationship with the actual positional deviation amount of the position reference mark. Therefore, the amount of misalignment of the mark can be obtained by obtaining the separation distance of the feature points and multiplying it by a constant. The constant here can be obtained from values measured several times at different positions.

  Alternatively, the amount of displacement can be detected based on the size of the region including the two feature points. When the feature points are very close or have a positional relationship such that some or all of them overlap, the size S of the feature point region and the positional deviation amount L have the following relationship.

  L = A × S + B

  Here, A and B are predetermined constants. These constants are designated by the size and positional relationship of the feature point area.

  By the way, the constant B in the above formula is half the average of the sizes of the two feature point regions. The constant A can be obtained by measuring L and S while changing the image positions of the two feature points. As the size of the feature point area, the minimum rectangular diagonal line or side length including the feature point area, the length of a line segment obtained by cutting a straight line connecting the centroids of the feature point area, or the like can be used. Therefore, as shown in FIG. 10, the amount of displacement can be detected based on the line segment C connecting the centroids 1 and 2 of the two feature points 151 and 161, respectively.

  Further, when there is one feature point region, the value of the constant A can be obtained by the method shown in FIG. The method for determining the constant B is to move the object so that there are two feature point regions and select a value that is half the average of the size of each feature point region, or select an arbitrary value empirically. May be.

  Furthermore, as shown in FIG. 11A, the region including the two feature points 151 ′ and 161 ′ may be a sum region of the feature point regions. Alternatively, as shown in FIG. 11B, an area including two feature points 151 ′ and 161 ′ may be an image area formed by the overlap of feature points. Furthermore, as shown in FIG. 11C, a minimum quadrangle including two feature points 151 ′ and 161 ′ may be used.

  That is, it is easy to determine whether there are two feature point regions by setting a region including two feature points. When two feature points are separated from each other, a region including them is large. This region can have various shapes, but in the case of image processing or the like, a rectangular shape is preferable because the coordinate values on the image can be used as they are.

  In addition, the amount of displacement can be detected by recognizing the shapes of the two feature points. That is, when the feature point has a shape that can be easily detected using the shape recognition method, the amount of positional deviation can be detected from the shape after the shape is detected.

  As shown in FIG. 12, the procedure first derives a line (step S41) and detects the shape (step S42). Next, the center of gravity / center of the shape is obtained (step S43), and the amount of displacement is calculated from the obtained center of gravity / center (step S44).

  As a shape recognition method, a Hough transform, a least square method, a template matching method, or the like can be employed. FIG. 13 shows a procedure when the Hough transform is employed. First, a line is detected (step S51), a circle is detected by Hough transform (step S52), and the amount of displacement is calculated from the center coordinates (step S53).

  The Hough transform can detect shapes (straight lines, circles, ellipses, etc.) that can be expressed by mathematical formulas, and can simultaneously determine the center and radius of the detected circles. Therefore, it is convenient that the positional deviation direction can be obtained simultaneously by the size and the positional deviation amount can be obtained by the distance between the centers.

  The least square method or the template matching method can detect a complex shape or a region having the closest brightness / color tendency in the shape. For example, when one of the position reference marks is a rectangle and the other is a circle, a feature point region can be easily specified by searching for a rectangle in the image by a template matching method and searching for a circle by Hough transform. And the shift direction of a feature point can also be detected simultaneously. In the procedure shown in FIG. 12, when the least square method or the template matching method is adopted, the derivation of the line in step S41 is not necessary.

  As shown in FIG. 14A, when the feature points 151 ′ and 161 ′ have a circular shape or a shape including a circular shape when viewed from the observation direction, the circles 151 ″ and 161 ″ are detected, and the distance between the centers thereof is detected. Based on this, it is possible to easily detect the amount of displacement. In addition, as shown in FIG. 14B, when both feature points 151 ′ and 161 ′ are circular when viewed from the observation direction, when the positional relationship is such that part or all of them overlap, The points 151 ′ and 161 ′ are of a gourd type. When observed as a gourd-shaped feature point, the feature points 151 ′ and 161 ′ can be determined from the radii of the circles 151 ″ and 161 ″.

  In the first embodiment, the position of the object to be adjusted or the observation position may be adjusted so that the two observed feature points have different sizes. In the case of observation with the shape of the object to be adjusted, the installation position, the observation position, and the optical system, the size of the feature point region changes depending on the focal length, angular magnification, and the like of the observation optical system. By using this, the position of the object to be adjusted, the observation position, and the focal length of the optical system are changed so that the two feature points have different sizes, thereby making it easy to detect the misalignment direction.

  Also, the installation position of the object to be adjusted, the observation position, or the focal length of the observation optical system may be adjusted so that the brightness of the two observed feature points is different, or the brightness of the illumination means is adjusted. May be.

  Furthermore, in the first embodiment, it is preferable that the shapes of the two observed feature points are different from each other when viewed from the observation direction, which facilitates the extraction of the feature point region and the determination of the displacement direction. Further, as shown in FIG. 15, at least one of the position reference marks may be a concave portion 171 or a convex portion 172 formed on the object 170, and at least one of the position reference marks is printed on the object 170. It may be what was done. Further, at least one of the position reference marks may be a cavity 173 formed inside the object 170.

(Refer 2nd Example and FIGS. 16-18)
In the positional deviation detection method according to the present invention, the inclination of the object can be detected by detecting two feature points. Here, an inclination adjusting apparatus using this detection method will be described as a second embodiment.

  As shown in FIG. 16, this tilt adjustment device is provided with an object 135 to be adjusted on the adjustment table 130, and the adjustment table 130 is driven to rotate about the X axis by the X axis rotation drive mechanism 141. The to-be-adjusted object 135 is driven to rotate about the Y axis by the attached Y axis rotation drive mechanism 142.

  The tilt of the object to be adjusted 135 is adjusted by the drive mechanisms 141 and 142 around the X and Y axes. Further, as shown in FIG. 17, a position reference mark 136 printed in an annular shape is provided on the back surface of the object 135 to be adjusted, and the position reference mark 137 printed in a circular shape having a smaller radius than the mark 136 is provided on the front surface. Is provided.

  The observation apparatus (CCD camera) 100 is the same as that shown in the first embodiment. The marks 136 and 137 are photographed, the photographed feature points are image-processed to calculate the shift amount, and The inclination of the object to be adjusted 135 is obtained. Then, the drive mechanisms 141 and 142 are driven based on the obtained inclination to adjust the inclination of the article 135 to be adjusted.

  The observation apparatus 100 transfers an 8-bit grayscale image of 1024 × 1280 pixels to the CPU 105 provided with the display 106. The CPU 105 corrects the tilt of the object 135 to be adjusted according to the procedure shown in FIG.

  The parameter Dx here is the radius of the observed feature point and the sensitivity of the rotation angle with respect to the pixel. The radius of each feature point in the image is measured in advance in units of pixels. The parameter Dx is obtained by measuring the deviation of feature points when the X axis is rotated by 0.1 ° in units of pixels. In addition, you may obtain | require and use the sensitivity regarding a Y-axis, and a unit is arbitrary.

  That is, the CPU 105 obtains the input image (step S61), detects the edge of the feature point area of the marks 136 and 137 (step S62), and further detects the feature point circle by Hough transform (step S63). Then, a circle closest to the radius of each feature point is searched from the detected circles (step S64).

  Next, a coordinate difference V between the centers of the two searched circles is obtained (step S65), and (θx, θy) · 10DxV is calculated (step S66). Then, it is determined whether or not the angle deviation amount | (θx, θy) | is smaller than the required accuracy t (step S67). If it is smaller, the operation is terminated. If larger, the process returns to step S61.

  Further, in order to calculate the inclination from the deviation amount of the marks 136 and 137, when the apparent positional deviation amount of the two feature points is x and the distance between the two feature points is r, the deviation angle is calculated. θ can be obtained using the following equation.

θ = sin ⁻¹ (x / r)

  When the distance r is sufficiently large or the positional deviation amount x is sufficiently small, x / r may be approximated and used.

  When deviation / tilt correction is performed using an XY table driving unit, a tilt correcting unit, or the like, it is necessary to perform correction processing many times in order to obtain a desired accuracy due to a driving error. However, in the second embodiment, the correction amount can be calculated as needed while observing the marks 136 and 137 of the adjustment target 135 that is the correction target, so that the deviation and inclination can be corrected efficiently.

(Position displacement detection on the front and back sides)
Further, as shown in FIG. 6B, by detecting the two marks 171 and 172 as feature points, a relative positional deviation between the front and back surfaces of one object 170 can be detected. In this case, since the misregistration can be calculated from one image at high speed, the misregistration detection can be speeded up and applied to an automatic inspection machine or the like.

(Refer to other tilt adjustment methods in the second embodiment, FIG. 19 and FIG. 20)
Next, another inclination adjusting method using the inclination adjusting apparatus shown as the second embodiment will be described. Here, as shown in FIG. 19, a position reference mark 138 printed in a cross shape is provided on the back surface of the object to be adjusted 135, and the position printed in a circular shape with a size that does not completely hide the mark 138 on the front surface. A reference mark 139 is provided.

  The observation apparatus 100 transfers an 8-bit grayscale image of 1024 × 1280 pixels to the CPU 105. The CPU 105 corrects the inclination of the object 135 to be adjusted according to the procedure shown in FIG.

  The parameter Dx here is the radius of the observed feature point and the sensitivity of the rotation angle with respect to the pixel. The radius of each feature point in the image is measured in advance in units of pixels. The parameter Dx is obtained by measuring the shift of the feature point when the X axis is rotated by 0.1 ° in units of pixels, and may be used by obtaining the sensitivity with respect to the Y axis.

  That is, the CPU 105 acquires the input image (step S71), detects the edge of the feature point area of the marks 138 and 139 (step S72), and applies the circle Hough transform and the straight line Hough transform simultaneously to mark the mark. 138 and 139 are detected as feature points (step S73).

  Next, a coordinate difference V between the detected circle center and the detected intersection of the two straight lines is obtained (step S74), and (θx, θy) · 10DxV is calculated (step S75). Then, it is determined whether or not the angle deviation amount | (θx, θy) | is smaller than the required accuracy t (step S76). If it is smaller, the operation is terminated. If larger, the process returns to step S71.

  Note that the method of calculating the inclination from the deviation amount of the marks 138 and 139 is as described in the second embodiment.

(Refer to another example of the position reference mark, FIG. 21)
In addition to the shapes described above, various shapes can be used as the position reference mark. For example, a rectangular mark 145 shown in FIG. 21A, a rectangular mark 145 arranged in a cross shape as shown in FIG. 21B, a rhombus mark 146 shown in FIG. 21C, or shown in FIG. An elliptical mark 147 may be used. Of course, shapes and arrangements other than these can be used as position reference marks.

(Characteristic point superposition procedure, see FIGS. 22 to 33)
Here, various procedures for superimposing the two feature points when adjusting the positions of the two objects 150 and 160 shown in FIG. 6A as the third embodiment will be described.

  First, a first procedure for superimposing the feature points A1 and A2 will be described with reference to FIGS. FIG. 22 shows a state where two position reference marks are observed as feature points A1 and A2 as an original image, and FIG. 23 shows a procedure for superimposing feature points A1 and A2.

  In the first procedure, first, the CPU acquires an original image (step S101). When the images do not overlap (see FIG. 22A), two feature points A1 and A2 are observed. Is calculated (step S102), and the distance between the centroids is calculated. The distance is calculated in advance by measuring the distance in the screen using a length reference standard and the like, calculating the length per pixel, and multiplying the obtained inter-pixel distance of the center of gravity.

  Next, one lens L1 is moved using the driving device 30, and is accurately superimposed on the other lens L2 (step S103). In this case, if the lens L1 can be accurately moved, the feature points A1 and A2 overlap (see FIG. 22C), and the light intensity distribution on the X and Y axes is symmetric. That is, the distribution in the X-axis direction and the Y-axis direction around the center of gravity value is calculated (step S104). If the distribution is bilaterally symmetric (YES in step S105), the process ends.

  However, when the accuracy of the driving device 30 is insufficient, the feature points A1 and A2 do not completely overlap (see FIG. 22B), and the light intensity distribution is asymmetrical (NO in step S105). Since the feature point A1 with the smaller light intensity becomes a form for imparting the light intensity to the feature point A2 with the larger light intensity, it can be said that the feature point A1 with the smaller base of the light intensity distribution is located. Therefore, when the light intensities of the feature points A1 and A2 are measured in advance, and when they overlap, what shape distribution is obtained based on the characteristics of the incident light intensity and the output value of the CCD image pickup device 13? And the distance between the feature points A1 and A2 can be calculated. The lens L1 is moved according to the calculated distance (step S106), and the feature points A1 and A2 are overlapped. Such an operation is repeated, and the above-described movement is performed until the light intensity distribution is completely left-right symmetric or until a certain symmetry is obtained.

  Next, a second procedure for superimposing the feature points A1 and A2 will be described with reference to FIGS. In the second procedure, first, the CPU acquires an original image (step S111), and performs boundary processing between the two feature points A1 and A2 (step S112).

  As shown in FIG. 24B, the boundary processing means the widths X2, Y2 (first square) occupied by the feature points A1, A2 on the X axis and Y axis, and X1, Y2 (second square). ) And its coordinates. Then, the larger feature point A2 is specified (step S113). Next, the entire boundary process including the feature points A1 and A2 is performed (step S114). That is, as shown in FIG. 24C, the widths X3 and Y3 of the entire rectangle and the coordinates thereof are calculated.

  If the two feature points A1 and A2 are completely overlapped, the result of the boundary processing of the larger feature point A2 is the same as the result of the entire boundary processing. The lens L1 is moved so that the result of the boundary processing of the larger feature point A2 is the same as the result of the entire boundary processing (steps S115 and S116).

  Next, a third procedure for superimposing the feature points A1 and A2 will be described with reference to FIGS. In the third procedure, the defect of the second procedure due to the boundary process is improved by performing a threshold process. The processing from step S111 to step S114 is similar to the second procedure.

  When the feature point A1 approaches the feature point A2 (see FIG. 26D), the light intensity between the feature points A1 and A2 overlaps to increase the light intensity, so that the total number of images after boundary processing is reduced. It may increase. In this case, it is difficult to determine which image is the feature point A1 or A2.

  Therefore, the threshold value of the light intensity is increased to calculate the number n of independent regions (step S114a). If the number n is 2 or less (YES in step S114b), the lens L1 is moved to determine the larger feature point A2. The lens L1 is moved until the boundary processing result is the same as the entire boundary processing result (steps S115 and S116). If the number n exceeds 2, the threshold value is further increased (step S114c), and the process returns to step S114.

  Next, a fourth procedure for superimposing the feature points A1 and A2 will be described with reference to FIGS. In the fourth procedure, in place of the threshold processing which is the third procedure, at least one feature point is moved to escape from the state in which the number of images is increasing, and the number of images after boundary processing is reduced by two. I am doing so. The processes from step S111 to step S114, steps S114a and 114b and the processes in steps S115 and S116 are the same as in the third procedure.

  When the threshold value of light intensity is increased and the number n of independent regions is calculated, if the number n exceeds 2 (NO in step S114b), one of the feature points is moved in a random direction (step S114d). ), The process returns to step S114. In this case, it is effective to move one feature point closer to the other feature point. This is because the light intensity at the overlapping portion of the two feature points A1 and A2 increases, and the number of images tends to become one by threshold processing.

  Next, a fifth procedure for superimposing the feature points A1 and A2 will be described with reference to FIGS. In the fifth procedure, the adjustment movement can be performed more effectively by using the third and fourth procedures together. The processes in steps S111 to S114, steps S114a and 114b, and the processes in steps S115 and S116 are the same as those in the third and fourth procedures.

  When the number n of independent regions is calculated by increasing the threshold value of the light intensity, if the number n exceeds 2 (NO in step S114b), the threshold value is further increased and one of the feature points is moved randomly. (Step S114e), the process returns to Step S114.

  Next, a sixth procedure for superimposing the feature points A1 and A2 will be described with reference to FIGS. The overlap image of the feature points A1 and A2 obtained in the fifth procedure becomes a circle when completely overlapped, and becomes an ellipse when the overlap is incomplete. Therefore, in the sixth procedure, the inscribed circle of the overlapping images is calculated. That is, the processes in steps S111, S112, and S114 are the same as those in the third procedure, but step S113 is not processed.

  When the feature point A1 approaches the feature point A2 (see FIG. 32D), the light intensity threshold value is raised to calculate the number n of independent regions (step S114a). (YES in step S114f), the lens L1 is moved (step S115), and the inscribed circle diameter φ of the overlapping image is calculated (step S117). When the feature points A1 and A2 are completely overlapped, the diameter of the inscribed circle is maximized. Therefore, until the feature points A1 and A2 are maximized or until the diameter φ of the inscribed circle becomes equal to the predetermined value φ0 (YES in step S118) Steps S115, S117, and S118 are repeated.

  On the other hand, when the number n of independent regions is calculated by increasing the threshold value of the light intensity (step S114a), if the number n is not 2 (NO in step S114f), one of the feature points is set at a fixed distance in the random direction. Move (step S114g) and return to step S114a.

(Other examples)
The position deviation detecting how according to the present invention is not limited to the embodiments can be modified in various ways within the scope of the invention.

  For example, as the image pickup device, various devices can be used as long as they can convert light energy into electrical output, such as CMOS, in addition to the CCD.

  In the present invention, the object to be adjusted is a transparent optical element, a transparent substrate, or a member having a hole. However, even if the object is opaque, the present invention can be applied by illuminating light rays (for example, ultraviolet rays) that pass through the object.

It is a schematic block diagram which shows the position adjustment apparatus using 1st Example of this invention. It is a flowchart figure which shows the basic adjustment procedure in 1st Example. It is a flowchart figure which shows the parameter setting procedure in 1st Example. It is a flowchart figure which shows the procedure which correct | amends the position shift in 1st Example. It is a chart figure which shows the frequency distribution of the brightness of the observed feature point. It is a schematic diagram which shows the position shift detection which can be implemented by this invention. It is explanatory drawing which shows the observation aspect by the position reference mark formed in the to-be-adjusted object. It is explanatory drawing which shows distribution of the brightness when the observed feature point overlaps. The observed characteristic point region and the surrounding brightness bias are shown, (A) is an explanatory diagram, and (B) is a chart. It is explanatory drawing for calculating | requiring the positional offset amount of the observed feature point. It is explanatory drawing which shows the area | region of the two feature points observed. It is a flowchart figure which shows the procedure which calculates | requires positional offset amount based on the shape of the observed feature point. It is a flowchart figure which shows the procedure which calculates | requires the positional offset amount by Hough conversion. It is explanatory drawing for calculating | requiring a positional offset amount by detecting a circle | round | yen. It is sectional drawing of the to-be-adjusted object which shows the aspect of a position reference mark. It is a schematic block diagram which shows the inclination adjustment apparatus using 2nd Example of this invention. The to-be-adjusted object in 2nd Example is shown, (A) is a front view, (B) is a bottom view. It is a flowchart figure which shows the procedure which adjusts inclination in 2nd Example. The to-be-adjusted object which has the position reference mark of the other shape used in 2nd Example is shown, (A) is a front view, (B) is a bottom view. It is a flowchart figure which shows the other procedure which adjusts inclination in 2nd Example. It is explanatory drawing which shows the position reference mark of another shape and arrangement | positioning. It is explanatory drawing which shows the 1st procedure which superimposes a feature point. It is a flowchart figure which shows the 1st procedure which superimposes a feature point. It is explanatory drawing which shows the 2nd procedure which superimposes a feature point. It is a flowchart figure which shows the 2nd procedure which superimposes a feature point. It is explanatory drawing which shows the 3rd procedure which superimposes a feature point. It is a flowchart figure which shows the 3rd procedure which superimposes a feature point. It is explanatory drawing which shows the 4th procedure which superimposes a feature point. It is a flowchart figure which shows the 4th procedure which superimposes a feature point. It is explanatory drawing which shows the 5th procedure which superimposes a feature point. It is a flowchart figure which shows the 5th procedure which superimposes a feature point. It is explanatory drawing which shows the 6th procedure which superimposes a feature point. It is a flowchart figure which shows the 6th procedure which superimposes a feature point.

Explanation of symbols

A1, A2 ... feature point 100 ... observation device 105 ... CPU
107 ... XY drive stage 110 ... Adjusted object 111, 121 ... Position reference mark 120 ... Adjusting member 130 ... Adjustment table 135 ... Adjusted object 136,137,138,139 ... Position reference mark 141,142 ... X axis, Y Shaft rotation drive mechanism 145, 146, 147 ... Position reference mark 150, 160, 170, 180 ... Object 151, 161, 171, 172, 181, 182 ... Position reference mark

Claims (16)

A displacement detection method for detecting a displacement amount and / or displacement direction of a position reference mark from a predetermined position based on the density of an image observed using a position reference mark formed on an object as a feature point ,
Two position reference marks are set, and one position reference mark is set as a predetermined position to detect a position shift amount and / or a position shift direction of the other position reference mark;
A misregistration detection method characterized by the above. The position shift detection method according to claim 1 , wherein a position shift direction and / or a position shift amount is detected on the basis of a region including the two feature points and / or shading around the region. Based on the overlap image that appears when the first feature point overlaps part or all of the second feature point as viewed from the observation direction and / or the brightness deviation around the overlap image, and / or The positional deviation detection method according to claim 2 , wherein a positional deviation amount is detected. The first feature point is viewed from the viewing direction a second portion of the feature points or the whole and overlapping positional deviation direction and on the basis of the gradient of the brightness around the overlap image and / or heavy becomes image appears when / Alternatively, the positional deviation detection method according to claim 2 , wherein the positional deviation amount is detected. The position shift detection method according to claim 2 , wherein the region including the two feature points is a sum region of the regions of the two feature points. 3. The positional deviation detection method according to claim 2 , wherein the region including the two feature points is an image region formed by overlapping the first feature point with the image of the second feature point. The position shift detection method according to claim 2 , wherein the region including the two feature points is a minimum quadrangle including the two feature points. The position shift detection method according to claim 1 , wherein a position shift direction and / or a position shift amount is detected based on a size of a region including the two feature points. The misregistration detection method according to claim 8 , wherein the misregistration amount is detected based on a line segment formed by a cross section obtained by cutting the region including the two feature points in the misregistration direction. The position shift detection method according to claim 1 , wherein a position shift direction and / or a position shift amount is detected based on the size of the observed two feature points. The position shift detection method according to claim 1 , wherein a position shift direction and / or a position shift amount is detected based on a difference in brightness between the observed two feature points. The position shift detection method according to claim 1 , wherein a position shift direction and / or a position shift amount is detected from a position difference between the two feature points. The position shift detection method according to claim 12 , wherein a position shift direction and / or a position shift amount is detected based on a line segment connecting the centroids of the two feature points. The positional deviation detection method according to claim 1 , wherein the positional deviation direction and / or the positional deviation amount are detected by recognizing shapes of the two feature points. 15. The positional deviation according to claim 14 , wherein a center of each of the two feature point regions is obtained using a detection method for detecting a circle, and a positional deviation amount is detected based on a distance between the centers. Detection method. The position shift detection method according to claim 14 or 15 , wherein the detection method for recognizing and detecting the shape is a Hough transform.