JP2000207557A - Method for measuring quantity of positional deviation - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce processing time by smoothing a virtually transformed image, specifying a coordinates point having a maximum count value in the smoothed virtually transformed image and defining the coordinates of the specified coordinates point as the positional deviation of a measurement object. SOLUTION: The relative coordinates between both edge points are calculated about the combinations of respective all of some tens of edge pints of an obtained reference image with respective all of some hundreds of edge points of an edge extracted image, and the calculated relative coordinates are plotted on a virtual plane 7. In such a case, the count value of coordinates points (x' and y') to be plotted is counted up by one whenever plotting to obtain a virtually transformed image. The virtually transformed image is smoothed. Scanning is performed in the smoothed virtually transformed image 9, coordinates point A having a maximum count value Fmax is specified, and the X and Y components (x'max and y'max) of the specified coordinates point A are defined as the positional deviations (x'max and y'max) of a measurement object. Thus, the positional deviation measurement of one piece of a measurement object is finished.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a method for measuring a displacement amount used for accurately positioning an object to be measured such as a circuit board or a wafer substrate in a planar direction.

[0002]

2. Description of the Related Art A method using a template matching method of image processing as a method of measuring a displacement amount used for positioning an object to be measured such as a circuit board having an image pattern formed on its surface by image processing. Is conventionally known. In this method, as shown in FIGS. 1, 11, and 12, a reference image 2 serving as a reference for pattern matching is registered in a memory as a template 10, and the template 10 is stored on an input image 1 of a measurement target product. The template 10 and the input image 1 are collated while being raster-scanned, and a position in the input image 1 where a pattern substantially matching the template 10 is found. As the template 10, an image of a part whose pattern is characteristic and which can be distinguished from other parts and whose position can be specified is selected from all the images of the reference target product serving as a model.

An example of a conventional method of measuring the amount of displacement is to first capture an image of a reference range of a reference target product, and include, for example, digital data (for example, density data) held at each pixel point of 32 pixels by 32 pixels. Is registered in the memory as a template 10 to be used. At the time of measurement, the target range of the measurement target product is imaged, and an input image 1 composed of digital data of each pixel point of, for example, 500 pixels horizontally × 500 pixels vertically is stored in a memory. Next, the template 10 is raster-scanned by software on the input image 1 and the degree of matching between the template 10 and the input image 1 in a portion where the template 10 overlaps is shifted by one dot each time by the following template residual. The position of the object to be measured is detected based on the scanning position at which the evaluation is made based on the difference R and the template residual R is minimized.

The reference point (i = 0, j =
When the scanning position of (0) is at a certain coordinate (x, y) on the input image 1, the template residual R (x, y) is obtained by inputting the value T of all points of the template 10 and overlapping each point. The sum of the absolute value of the difference from the value P of the pixel point on the image 1 is expressed by the following equation.
Calculate in (1).

[0005]

(Equation 1)

Here, (i, j) represents the coordinates on the template 10, T (i, j) represents the data value of the template 10 at the coordinates (i, j) on the template 10, P
(X, y) indicates the data value of the input image 1 at the coordinates (x, y) on the input image 1, respectively.

[0007]

However, in the conventional method for measuring the amount of displacement, as can be seen from equation (1), the data values T (i, j) of all points in the template 10 correspond to the corresponding input values. Pixel point value P (x, y) on image 1
If the value of the template residual R does not become sufficiently close to the value, the template residual R does not become a sufficiently small value. That is, if the reflectance of the whole or part of the measurement target varies, or if the illuminance of the whole or part of the measurement target fluctuates due to illumination, even if the template 10 is at the scanning position where the pattern matches, the template residual R Is not a sufficiently small value. In other words, if there is dirt such as fogging on the surface of the object to be measured or illumination unevenness, the scanning position where the pattern almost matches may not be detected due to such influences, and the reliability of the measurement of the displacement amount may be reduced. There is a problem that it is not high.

Further, the template residual R (x,
Since the calculation of y) is performed for almost all the coordinate points on the input image 1, there is a problem that an enormous amount of calculation processing is required for measuring the amount of displacement of one circuit board, and a long processing time is required. That is, even if only the subtraction process of PT is performed, the number of calculation processes becomes a huge number (several hundred millions) of the product of the total number of pixels of the input image 1 and the total number of pixels of the reference image 2.

SUMMARY OF THE INVENTION In view of the above problems, the present invention makes it possible to measure a position shift amount with high reliability, which is hardly affected by dirt and uneven illumination of a measurement object, and to reduce the processing time. It is an object of the present invention to provide a method for measuring a displacement amount that can be performed.

[0010]

SUMMARY OF THE INVENTION In order to achieve the above object, the present invention provides an object to be measured based on an input image obtained by imaging the object to be measured and a reference image serving as a reference for image comparison and position. In the displacement amount measurement method for calculating the displacement amount of the reference edge point is extracted from the reference image, a master image including all the extracted reference edge points is registered, and the target range of the measurement target product is imaged. An edge point is extracted from the captured input image, an edge extraction image including all the extracted edge points is stored, and both the reference edge points of the master image and the edge points of the edge extraction image are combined. Each time the relative coordinates between the edge points are obtained and the obtained relative coordinates are plotted on the virtual plane, the count value of the plotted coordinate points is incremented by one, and the virtual converted image is obtained. After that, the virtual conversion image is smoothed, the coordinate point having the largest count value is specified in the smoothed virtual conversion image, and the coordinates of the specified coordinate point are used as the positional deviation amount of the measurement object. Features.

According to the displacement measuring method of the present invention, the following operation can be performed. First, plotting the relative coordinates between a certain reference edge point of the master image and each of all the edge points of the edge extracted image on a virtual plane, a plot point by one edge point matching the one reference edge point is obtained. A certain plot point group based on all edge point groups of the edge extraction image including the distribution is distributed on the virtual plane (see plot point group 8a in FIG. 6).

Next, a group of plot points different from each other are distributed on the virtual plane for each of the other reference edge points of the master image (the plot point group 8 in FIG. 6).
b, 8c). Now, in the group of many different plot points, there is one plot point by one edge point that conforms to each reference edge point, and all the plot points are the master image and the edge extracted. The coordinates should be substantially the same as the coordinates relative to the image (see point A in FIGS. 6 and 7). Therefore, for each combination of all the reference edge points of the master image and all the edge points of the edge extracted image, the relative coordinates between the two edge points are obtained, and the plotted coordinates are plotted every time the obtained relative coordinates are plotted on the virtual plane. When the count value of a point is counted up by one to obtain a virtual conversion image, a large number of different plot points are superimposed on the virtual plane, so that the relative coordinate points of the master image and the edge extraction image and The count value in the vicinity becomes a large value, or the plot points are concentrated near the relative coordinate points (see point A in FIGS. 7 and 8). Then, when the virtual conversion image is appropriately smoothed, the relative coordinate point between the master image and the edge extracted image or a point in the immediate vicinity thereof should show the maximum count value (see point A in FIG. 10). Accordingly, the coordinate point A having the maximum count value Fmax is specified in the smoothed virtual conversion image, and the coordinates (x'max, y'max) of the specified coordinate point A are used to determine the positional deviation amount (x 'max, y'max) (see FIG. 10). The total number of combinations of all the reference edge points of the master image and all the edge points of the edge extracted image is an order of the order of the product of the number of horizontal and vertical pixels of the master image and the number of horizontal and vertical pixels of the input image. As a result, compared with the conventional template matching method, the amount of displacement can be measured with a much smaller number of calculation processes, and the processing time can be reduced.

Further, since the edge point is detected from the input image of the object to be measured, it is hardly affected by a change in the absolute value of a value at a certain point of the input image. And a highly reliable displacement amount measurement that is hardly affected by the above.

The master image is obtained by capturing a reference range of a reference object as a model, taking in as a reference image in which each pixel point distributed in a matrix has digital data, and for each pixel point in the matrix. The digital data is differentiated at least in one direction of the matrix, and a pixel point whose absolute value of the differential value obtained for each pixel point is larger than the absolute value of the differential value of both adjacent pixel points in at least one direction is defined as a reference edge. An edge-extracted image is obtained by imaging the target range of the measurement target product, and each pixel point distributed in a matrix is captured as an input image having digital data. The digital data is differentiated in at least one direction of the matrix for each pixel point of the absolute value of the differential value of the differential value of both adjacent pixel points in at least one direction If the pixel points larger than the absolute value are configured as all the edge points extracted as edge points, only the inflection points of the curve connecting the digital data in the differential direction are effectively extracted as edge points. The number of edge points to be extracted can be minimized, and the amount of displacement can be measured with less computational processing. In addition, since both the master image and the edge-extracted image can be obtained by processing the image obtained by imaging the target product in the same manner, the master image and the edge-extracted image can be obtained under the same imaging conditions and processing procedures. In addition, the degree of coincidence of the corresponding portion between the master image and the edge extraction image can be increased, and more reliable measurement of the amount of displacement can be performed.

For each combination of all the reference edge points of the master image and all the edge points of the edge extracted image, relative coordinates between the two edge points are obtained, and the obtained relative coordinates are plotted every time they are plotted on a virtual plane. If the count value of the coordinate point is configured to be counted up based on the absolute value of the differential value of the edge point of the edge extracted image corresponding to the coordinate point to obtain a virtual conversion image, a point plotted on a virtual plane Can be counted up in consideration of the magnitude of the absolute value of the differential value of the corresponding edge point, so that sharp edge points can be extracted more conspicuously. A more reliable measurement of the amount of displacement can be performed with less influence of illumination unevenness.

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiments of the present invention will be described below with reference to the drawings.

As shown in FIGS. 1 to 10, an embodiment of the method for measuring the amount of displacement according to the present invention is to obtain an input image 1 obtained by capturing an object to be measured (circuit board) in a first step by extracting edges. The extracted edge-extracted image 4 and the master image 6 obtained by edge-extracting the reference image 2 serving as a reference for image collation and position in the second step are subjected to image processing in the subsequent third to fifth steps to obtain a final image. (X'max, y'ma
x). The steps will be described below.

In the first step, first, a reference range of a reference object (a good circuit board serving as a model) is imaged by a CCD camera, and each pixel point distributed in a matrix as shown in FIGS. (I, j) is captured as a reference image 2 having digital data T (i, j).

Next, for each pixel point (i, j) of the matrix, the digital data T (i, j) is divided into a horizontal differential operator H (u, v) shown in FIG. 2 and a vertical differential operation shown in FIG. Child V
Differentiate by (u, v). The specific calculation procedure is
A 3 × 3 dot horizontal differential operator H (u, v) shown in FIG. 2 is superimposed at a certain coordinate (i, j) on the reference image 2, and each of nine points of the horizontal differential operator H (u, v) is The product of the dot and the data value of the corresponding point of the reference image 2 is represented by the horizontal differential operator H (u,
The differential value D summed (convolution operation) in v)
_i (i, j) is obtained, and similarly, a differential value D _j (i, j) obtained by performing a convolution operation using the vertical differential operator V (u, v) of 3 × 3 dots shown in FIG. 3 is obtained.

Next, the two differential values D _i (i, j), D
_j The sum S (i, j) of the absolute values of
(I, j) = | D _i (i, j) | + | D _j (i, j) |
Is calculated as

Here,

[0022]

(Equation 2)

Next, a point (i, j) where S (i, j) is larger than both adjacent points in the horizontal direction X or the vertical direction Y by a predetermined threshold Q or more, ie, S (i, j). j) ≧ S (i−1, j) + Q and S (i, j) ≧
S (i + 1, j) + Q or S (i, j) ≧ S (i, j−1) + Q and S (i, j) ≧
All points (i, j) satisfying S (i, j + 1) + Q are extracted as reference edge points 5, and a master image 6 including all the extracted reference edge points 5 is registered in the memory. This master image 6
As shown in FIG. 4, reference edge points 5 which are edge elements of the reference image 2 are linearly distributed. The number of the reference edge points 5 depends on the complexity of the pattern of the reference image 2 and the like, but is an order of about the sum of the number of horizontal pixels and the number of vertical pixels of the reference image 2, and is several tens in the present embodiment. .

In the second step, first, an image of the target range of one measurement target product is taken, the image is taken with a CCD camera, and each pixel point (x, y) distributed in a matrix as shown in FIG. Capture as input image 1 having data P (x, y).

Next, the digital data P (x, y) for each pixel point (x, y) of the matrix is divided into the horizontal differential operator H (u, v) and the vertical differential operator V (u, v). ) To differentiate. A specific calculation procedure is to obtain a value D _x (x, y) obtained by performing a convolution operation using a horizontal differential operator H (u, v) at a certain coordinate (x, y) on the input image 1, and similarly calculate a vertical differential A value D _y (x, y) obtained by performing a convolution operation with the operator V (u, v) is obtained.

Next, the two differential values D _x (x, y), D
_y (x, y) The sum S (x, y) of the absolute values of
(X, y) = | D _x (x, y) | + | D _y (x, y) |
Is calculated as

Here,

[0028]

(Equation 3)

Next, a point (x, y) where S (x, y) is larger than both adjacent points in the horizontal direction X or the vertical direction by the predetermined threshold Q or more, ie, S (x, y). , Y) ≧ S (x−1, y) + Q and S (x, y) ≧
S (x + 1, y) + Q or S (x, y) ≧ S (x, y−1) + Q and S (x, y) ≧
All the points (x, y) satisfying S (x, y + 1) + Q are extracted as edge points, and an edge extracted image 4 including all the extracted edge points 3 is obtained as
Store in memory. The number of the edge points 3 is an order of about the sum of the number of horizontal pixels and the number of vertical pixels of the input image 1, and is several hundred in the present embodiment.

As shown in FIG. 4, the obtained edge extracted image 4 has the edge points 3 which are the edge elements of the input image 1 distributed linearly, and substantially matches the master image 6 described above. The portion should be displaced from the master image 6 by the amount of displacement finally desired, as indicated by the vector G in FIG. Hereinafter, the XY components of the vector G will be obtained by software.

In the third step, both edge points 3, 5 are determined for each combination of several tens of all reference edge points 5 of the obtained master image 6 and several hundreds of all edge points 3 of the edge extracted image 4. The relative coordinates between them are obtained, and the obtained relative coordinates are plotted on the virtual plane 7. At this time, every time the plotting is performed, the count value E of the coordinate point (x ′, y ′) to be plotted is calculated.
(x ′, y ′) is counted up by one to obtain a virtual conversion image 8 shown in FIG. The procedure of this third step will be described in more detail below.

First, as shown in FIG.
Extracted image 4 for one reference edge point 5 having
When the relative coordinates (XY components of the vectors K _1, K _{2, to} K _m ) of all the edge points 3 are plotted on the virtual plane 7, as shown in FIG. The same plot point group 8a as the edge extraction image 4 consisting of several hundred points is plotted.

Similarly, when the above procedure is performed for several tens of other reference edge points 5, as shown in FIGS. 6 and 7,
Each reference edge point 5 is defined as the origin O 'and the edge extracted image 4
The same plot point group (for example, 8b and 8c) is plotted on the virtual plane 7 while being slightly shifted from each other. Then, as shown in FIG. 8, each time one plot point group is plotted, if the plots overlap at each coordinate point (x ′, y ′), the count value of that coordinate point (x ′, y ′) E (x ', y') is counted up by one. For convenience of explanation, FIG. 8 shows a much smaller number of plot points (x ′, y ′) and a smaller count value E (x ′, y ′) than actual.

Assuming that the point A is a point having coordinates of the XY components of the vector G, each plot point group (for example, 8a, 8a
b, 8c), there is a point on the point A or in the vicinity of the point A by one edge point 3 corresponding to each reference edge point 5, so that all plot points are plotted as shown in FIG. As shown in FIG. 7, several tens of points are concentratedly plotted on or near the point A. Therefore, FIG.
As shown in the reference, the count value E on or near the point A becomes larger than the other parts.

Ideally, dozens of points overlap the point A on the virtual plane 7 to maximize the count value E of the point A. In practice, however, the reference image 2 of the input image 1 Since the pattern of the corresponding portion has a variation in the pattern shape of the measurement target product and an error at the time of imaging, it is rare that the pattern completely matches the pattern of the reference image 2. It is common that individual points are concentrated. in this case,
Since the count value E of the point A does not always become the maximum, it is more accurate to perform a local smoothing process to obtain the center point of the portion where the count value E is most concentrated, and use this as the point A. Becomes Therefore, before the XY component of the point A is determined, the following fourth step is further performed.

In the fourth step, the virtual conversion image 8 is smoothed by a smoothing operator B (u, v) shown in FIG. 9 for every plot point (x ', y') (FIG. 10). reference). The specific calculation procedure is the plotted point (x ′, y ′) on the virtual plane 7.
The value F (x ', y') of

[0037]

(Equation 4)

And smoothing. At this time, every time the smoothing is performed, the count value F (x ′, y ′) of the coordinate point (x ′, y ′) is counted up to obtain the smoothed virtual conversion image 9 shown in FIG. As a result, the count value E (x ′, y ′) of each plot point (x ′, y ′) of the virtual conversion image 8 is changed to a 3 × 3 point having a weight at the center point (x ′, y ′). Is obtained, and the above-mentioned smoothed virtual conversion image 9 is obtained.

In the fifth step, the coordinate point A having the maximum count value Fmax is specified by scanning the smoothed virtual conversion image 9 shown in FIG. 10, and the XY component (x'max) of the specified coordinate point A is specified.
, y'max) and the displacement (x'max,
y'max). Thus, the measurement of the displacement amount of one measurement target product is completed.

Regarding the displacement measurement of the second and subsequent measurement objects, the coordinate point A having the maximum count value Fmax is specified in accordance with the procedures of the second to fifth steps, and the displacement (x ′) is determined. max, y'max). The total number of combinations of each of the reference edge points 5 of the master image 6 and each of the edge points 3 of the edge extraction image 4 is of the order of several tens × several hundreds. Compared with this, the displacement amount can be measured with far less calculation processing.

In the above embodiment, the relative coordinates between the two edge points 3 and 5 are obtained for the combination of all the reference edge points 5 of the master image 6 and all the edge points 3 of the edge extracted image 4, and the obtained relative coordinates are obtained. Is plotted on the virtual plane 7, instead of incrementing the count value E (x ', y') of the plotted coordinate point (x ', y') by one each time plotting is performed (see FIG. 8). , And the count-up amount is defined as an edge point (x, y) corresponding to the coordinate point (x ′, y ′).
Of the absolute values of the differential values D _x (x, y) and D _y (x, y) of
Counted up as (x, y) and converted to virtual converted image 8
Is obtained, the edge point 3 of the sharp part can be more remarkably extracted than the edge point 3 of the other blurred part, and the measurement object is less susceptible to the effects of dirt and illumination unevenness and has higher reliability. The displacement amount can be measured.

In the above embodiment, to detect the reference edge point 5 and the edge point 3, the reference image 2 and the input image 1 are calculated by using the differential values D _i (i.e., calculated using the horizontal and vertical differential operators H and V). i, j), D _j (i, j), and D _x (x,
y), the sum S (i, i) of the absolute values of D _y (x, y)
Although j) and S (x, y) are detected to be points that are larger than both adjacent points in the horizontal direction X or the vertical direction Y, the present invention is not limited to this. A method may be used.

In the above-described embodiment, the reference object 2 serving as a model is imaged to obtain the reference image 2.
The reference image 2 may be obtained using design data such as D data.

[0044]

According to the displacement measuring method of the present invention, the following effects can be obtained. For each combination of all the reference edge points of the master image and all the edge points of the edge extracted image, the relative coordinates between the two edge points are obtained, and each time the obtained relative coordinates are plotted on the virtual plane, the coordinates of the plotted coordinate points are calculated. When the count value is incremented by one to obtain a virtual conversion image, the relative coordinate point between the master image and the edge extracted image and the count value in the vicinity thereof become a large value, or the plot points are concentrated near the relative coordinate point. I do. Then, when the virtual conversion image is appropriately smoothed, the relative coordinate point between the master image and the edge extraction image or a point in the immediate vicinity thereof indicates the maximum count value. Therefore,
The coordinate point having the maximum count value in the smoothed virtual conversion image can be specified, and the coordinates of the specified coordinate point can be used as the positional deviation amount of the measurement object.

Since the order is about the product of the number of horizontal and vertical pixels of the master image and the number of horizontal and vertical pixels of the input image, the displacement amount can be measured with a much smaller number of computations than the conventional template matching method. And the processing time can be reduced.

Further, since the edge point is detected from the input image of the object to be measured, it is hardly affected by a change in the absolute value of a value at a certain point in the input image, and even if the object to be measured is dirty or has uneven illumination. And a highly reliable displacement amount measurement that is hardly affected by the above.

The master image is obtained by differentiating the reference image obtained by imaging the reference target product in at least one direction for each pixel point,
It consists of all reference edge points obtained by extracting pixel points whose absolute value of the differential value obtained for each pixel point is greater than the absolute value of the differential value of both adjacent pixel points in at least one direction as reference edge points, The edge-extracted image is obtained by differentiating the input image obtained by imaging the object to be measured in at least one direction of the matrix for each pixel point, and the absolute value of the differential value is at least one direction of the differential value of both adjacent pixel points. If the pixel points larger than the absolute value are configured as all the edge points extracted as edge points, only the inflection points of the curve connecting the digital data in the differential direction are effectively extracted as edge points. The number of edge points to be extracted can be minimized, and the amount of displacement can be measured with less computational processing. In addition, since both the master image and the edge-extracted image can be obtained by processing the image obtained by imaging the target product in the same manner, the master image and the edge-extracted image can be obtained under the same imaging conditions and processing procedures. In addition, the degree of coincidence of the corresponding portion between the master image and the edge extraction image can be increased, and more reliable measurement of the amount of displacement can be performed.

Each time the relative coordinates between the two edge points are plotted on the virtual plane for each combination of all the reference edge points of the master image and all the edge points of the edge extracted image, the count value of the plotted coordinate points is calculated as follows: When configured to count up based on the absolute value of the differential value of the edge point of the edge extraction image corresponding to the coordinate point to obtain a virtual conversion image, sharp edge points can be more remarkably extracted, It is less susceptible to the effects of dirt and illumination unevenness on the object to be measured, and more reliable measurement of the amount of displacement can be performed.

[Brief description of the drawings]

FIG. 1 is a diagram showing an example of an input image and a reference image.

FIG. 2 is an enlarged view showing an example of a horizontal differential operator.

FIG. 3 is an enlarged view showing an example of a vertical differential operator.

FIG. 4 is a diagram showing a part of an edge extraction image and a master image.

FIG. 5 is a diagram showing a positional relationship between one reference edge point and an edge point.

FIG. 6 is a conceptual diagram illustrating superimposition of virtual conversion images.

FIG. 7 is a diagram showing a part of a virtual conversion image.

FIG. 8 is an enlarged view of a part of the virtual conversion image displaying the count value.

FIG. 9 is an enlarged view showing an example of a smoothing operator.

FIG. 10 is an enlarged view showing a part of a virtual conversion image in which a count value is smoothed.

FIG. 11 is a diagram showing an example of a template based on a reference image.

FIG. 12 is a conceptual diagram illustrating raster scanning on an input image.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Input image 2 Reference image 3 Edge point 4 Edge extraction image 5 Reference edge point 6 Master image 7 Virtual plane 8 Virtual conversion image 9 Smoothed virtual conversion image A Point with the required displacement amount as coordinates E Count of virtual conversion images Value F Count value of smoothed virtual converted image Fmax Maximum count value of smoothed virtual converted image P Digital data of input image T Digital data of reference image (template) X Horizontal Y Vertical x, y On input image Coordinates x ', y' Coordinates x'max, y'max on virtual plane

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 2F065 AA01 BB02 CC01 CC19 DD06 FF04 JJ03 JJ26 QQ31 RR06 5B057 AA03 BA02 CA02 CA08 CA12 CA16 CB02 CB08 CB12 CB16 CC01 CE03 CE05 CE06 DA07 DB02 DB05 DB09 DC03 DC16 DC32

Claims (3)

[Claims] 1. An input image obtained by imaging an object to be measured,
In a displacement amount measuring method for obtaining a displacement amount of an object to be measured based on an image collation and a reference image serving as a reference of a position, a reference edge point is extracted from the reference image, and a master including all the extracted reference edge points is extracted. An image is registered, an image of the target range of the measurement target product is captured, edge points are extracted from the captured input image, and an edge extraction image including all the extracted edge points is stored, and all reference edges of the master image are stored. For each combination of each point and all the edge points of the edge extracted image, the relative coordinates between the two edge points are obtained, and each time the obtained relative coordinates are plotted on the virtual plane, the count value of the plotted coordinate points is incremented by one. After counting up to obtain a virtual conversion image, the virtual conversion image is smoothed, and the coordinate point having the largest count value in the smoothed virtual conversion image is identified. A method for measuring the amount of positional deviation, wherein the coordinates of the specified coordinate point are used as the amount of positional deviation of the measurement object. 2. A master image is obtained by capturing an image of a reference range of a reference target product serving as a model, and capturing as a reference image in which each pixel point distributed in a matrix has digital data.
The digital data is differentiated at least in one direction of the matrix for each pixel point of the matrix, and the absolute value of the differential value obtained for each pixel point is the absolute value of the differential value of both adjacent pixel points in at least one direction. It consists of all the reference edge points that are extracted by extracting the larger pixel points as reference edge points.The edge extraction image captures the target range of the measurement target product, and each pixel point distributed in a matrix has digital data. The digital data is differentiated at least in one direction of the matrix for each pixel point of the matrix, and the absolute value of the differential value is calculated from the absolute value of the differential value of both adjacent pixel points in at least one direction. 2. The method according to claim 1, wherein all the edge points are obtained by extracting a large pixel point as an edge point. 3. For each combination of all reference edge points of a master image and all edge points of an edge extracted image, relative coordinates between the two edge points are obtained, and each time the obtained relative coordinates are plotted on a virtual plane, plotting is performed. The virtual conversion image is obtained by counting up the count value of the coordinate point to be obtained based on the absolute value of the differential value of the edge point of the edge extraction image corresponding to the coordinate point. How to measure displacement.