JP3067822B2 - Pattern identification method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a pattern identification method, and more particularly to a pattern separation and identification method for accurately and accurately identifying and measuring tableware in, for example, a cafeteria automatic payment system.

[0002] Here, the canteen automatic payment system refers to an image processing system in which when a variety of dishes or single dishes are put on a tray in a canteen, an image of a tableware including overlaps and protrusions is input to a camera and image processing is performed. This system calculates the settlement amount by comparing the size of tableware with the price of tableware determined in advance according to the processing result. In this case, it is necessary to accurately identify the pattern of the tableware image.

[0003]

2. Description of the Related Art Conventionally, in the identification and measurement of tableware, for example, Japanese Patent Application Laid-Open Nos. 61-147381 and 62-19 are disclosed.
Several methods are known in the invention proposed in US Pat. For example, there is a case in which what kind of circular tableware is formed from the pattern shape of the overlapping circular tableware is separated and identified, or a case in which the circular tableware and the polygonal object are separated and identified when they are overlapped. First, a case in which only the circular tableware shown in FIG. 16 is formed will be described.

FIG. 16A, which shows a binarized image of an object, is shown in FIG.
For a pattern as shown,
Gut points are extracted as shown in FIG. Of this contour
Find the inflection point as follows. First, as shown in FIG.
Equally spaced sampling points P ₁, P_Two, ..., P_m, ... P
_n, ... P_O, ... P_p,..., Then the adjacent 3
Vector V between two sampling points₁, V_Two(Figure
17 (B)) are sequentially obtained. Next, this one
Inflection point depending on whether the direction change value is within or outside the specified set value
P_m, P_n, P_O, P_pAsk for. Next, the found inflection
Between points, for example, inflection point P_m~ P_nAre the same circular dishes
Inflection point P_m, P _n, P_nm3
Consider a circle passing through a point and calculate the center coordinates and radius of the circle
From the calculation result, the radius and the center coordinates are close to each other
Things are integrated as the same circle. For example, figure
17 (A), two contour lines (P_m~ P_n) And (P
_o~ P_p) Are integrated as the same circle. More than
The process identified the type and number of the circles.

FIG. 18 shows a method of identifying a polygon.
This will be described with reference to FIG. First, the inflection points are determined as X ₁ , X ₂ , X ₃ , X ₄ , and X _{5 by} changing the direction of the inclination vector between the sampling points as in the case of the circle. Next, if the change in the direction of the gradient vector between the inflection points is a value close to 0, it is identified as a straight line and numbered (,,,), and then the interior angle with the next side of the target side (Q ₁ , Q ₂ ,
Q ₃ ) and the lengths of the sides are stored in a memory as shown in FIG. Next, the data stored in the memory and the inspection object are collated to identify which side of the polygon is the side of interest.

Next, description will be made on a case where separation and identification are performed when polygonal objects are overlapped. An outline point is extracted as shown in FIG. 20B by a contour tracing method for a pattern as shown in FIG. 20A in which the image of the object is binarized. As in the case described above, the concavo-convex points of this contour are shown in FIG.
As shown in FIG. 21A, the inflection point P _j changes as shown in FIG. 21A due to the change in the direction of the inclination vectors V ₁ and V ₂ (FIG. 21B) between the sampling points P ₁ and P ₂ . , P _k , P _l , P _m , P
_{Find n} , P _o , and P _p . Further, if the direction of the inclination vector is close to 0, it is identified as a straight line, and numbers (,,...,...) Are assigned between inflection points. Next, FIG. 22 shows the length and the like of the inner angle (for example, Q ₁ , Q ₂ , Q ₃ ) between the target side (for example, the side of the number) and the next side (for example, the side of the number). In the memory. Next, the data stored in the memory and the inspection object are collated to identify which side of the polygon is the side of interest.

[0007]

The conventional method includes, for example, a circle A, a square B, an irregular (arc) C, as shown in FIG.
If amorphous D is overlapping pattern, the overlapping part as irregular (arc) C is in contact with the circular A, difficult to find the inflection points, Therefore, the arc between the inflection point P ₁₁ to P ₁ k
Is recognized as existing. Therefore, FIG.
As shown in the figure, the operation (dashed line) on the arc k has a problem that an erroneous recognition occurs in the discrimination from the actual circle (solid line) A.
Further, there is a problem that it is difficult to identify the irregular shape C. Further, when the irregular shape D overlaps the rectangular shape B to form a polygon on the screen, even if the rectangular shape B exists, it is absorbed by the polygon and it becomes impossible to identify whether the desired shape is a rectangular shape or an irregular shape. Similarly, for example, a polygon H, as shown in FIG.
When I and J overlap, a polygonal pattern as shown in FIG. 23 (B) is formed on the screen. Therefore, the rectangle H is absorbed by the polygon and it is not possible to discriminate whether the desired shape is a rectangle or a polygon. There was a problem that became possible.

The present invention is directed to a composite shape pattern in which a circle and a rectangle are extracted even if an irregular shape is included in the composite shape pattern in which the circle and the rectangle are superimposed, and a plurality of rectangles and a complicated polygon are formed to overlap with each other. It is an object of the present invention to provide a means pattern identification method capable of extracting a quadrangle from a polygon and quickly and accurately identifying and measuring a desired shape.

[0009]

When an irregular shape is included in a composite shape pattern in which a circle and a rectangle are superimposed, when extracting the circle and the rectangle, in the circle, the angle formed by the partial arc contour line corresponds to the entire arc. and threshold set above for (condition 100), circular center coordinates separated identified by partial arc contours inflection point does not exist within a circular which is separately isolated identification (condition 101), variable The difference between the radius of the operation circle based on the center point of the operation circle calculated from the partial arc contour having no curved point and the distance from the center point to the arc contour line on the actual screen is equal to or smaller than a threshold (condition 102). By using the following equation, any circular shape is discriminated. If the circular discriminating condition (100 to 102) is not satisfied, all the shapes are excluded as irregular shapes. angle)
Is a right angle (condition 103), the vectors at the two opposing partial side contours are parallel and opposite in direction (condition 104), and there is an overlap between the two vectors (condition 105). ), Any square is identified, and if the above-described square identification conditions (103 to 105) are not satisfied, all are excluded as irregular shapes.

On the other hand, when extracting a quadrangle from a polygon in a composite shape pattern formed by overlapping a plurality of quadrangles and complicated polygons, if there are a plurality of partial side contours generated from the same side. For example, they are integrated into one straight line, and the vectors of the two opposing partial edge contours including the integrated partial edge contours are parallel and opposite in direction (condition 11).
0) and there is an overlap between the two partial edge contour vectors (condition 111), and there is a plurality of matching partial edge contour vectors in the target partial edge contour vector By selecting the nearest partial side contour vector (condition 112), two opposite vectors are extracted to identify a rectangle from a polygon, and if the above-described rectangle identification condition (110 to 112) is not satisfied, All are excluded as indefinite.

[0011]

[Function] When a plurality of types of circles and a shape pattern such as a square and an irregular shape are complicatedly overlapped, an irregular portion is excluded as a measurement target by using the circular discriminating condition and the square discriminating condition. Accurate and quick measurement of whether or not it is composed of circles and squares.

In the case of a polygonal pattern formed by overlapping a plurality of types of quadrangles and complex polygons, a quadrangle included in the polygon and another polygon are verified using a polygon identification method. . Other polygons are measured only for a quadrangle as being out of the measurement object, and quick measurement is performed.

[0013]

FIG. 2 is a block diagram showing a first embodiment of the present invention. In the figure, the shape pattern 1 to be inspected is a lighting device 2.
, And the image is taken by the image input device (camera) 3 to be a video signal. The output of the camera 3 is A / D converted by an A / D conversion circuit 4 to be image data, binarized by a logical filter processing circuit 5 and reduced in noise, and stored in an image memory 6. The image data read from the image memory 6 enters the measurement processing device 7, where the processing described below is performed. The measurement processing device 7 performs processing according to the flowchart shown in FIG.

Contour points are detected counterclockwise by a contour tracing method for the shape pattern as shown in FIG. 4A (step 50 in FIG. 3), and the contour points are chased to follow contour point coordinates (x, y) is extracted and vectorized (step 51),
It is stored in the contour point memory 8. FIG. 4B shows the directions of the angles formed by the vectors at the inflection points P ₁ and P ₄ . A vector connects adjacent contour points counterclockwise.
The angle formed by the vectors is calculated by defining ± directions as shown in FIG. 4B using the vectors V _m and V _{n of interest} and the vectors V _m−2 and V _n−2 immediately before, respectively. Step 5 of 3
2). When the angle formed by the vector suddenly becomes positive, the inflection point P ₁ is obtained, and when the angle formed by the vector suddenly becomes negative, the inflection point P ₄ is obtained.

However, FIG. 4 shows an embodiment, and the tracking of the contour line is also possible in the counterclockwise and clockwise directions.
It is also possible to define the sign by the sign of the angle formed by the vector.

In FIG. 4A, inflection points are P ₁ , P ₂ ,
... P _6, ..., to become P _11, the contour line between the inflection points (hereinafter referred to subpaths) a, b, ... e, ..., and k. next,
This partial outline is divided into equal parts as shown in FIG. The value of q is determined by the discrimination relative tolerance. Here, considering the resistance to small noise, angles α1, α2,..., Αq and β formed by the vector of interest and the vector two vectors ahead
, Β2,..., Βq and calculate the average value (α1 + α2
+... + Αq) / q and (β1 + β2 +... + Βq) / q. If this value is smaller than the permissible set value (close to 0), it is classified as a side contour generated from a polygon on the screen, and if larger, it is classified as a circular contour generated from a circle, as shown in FIG. The coordinates of the arc contours i and k included in the contours (a to k) are used as the arc contour parameter 9a, and
The coordinates of the side contours a, b, c, d, e, f, g, h, and i are stored in the side contour parameters 9b (step 5 in FIG. 3).
3) (However, in FIG. 6, x coordinates x _{a to} x _k and y coordinates y _a
Ｙy _k represents all coordinates of the partial contours a to _k ). Next, the circular and square extracting unit 10 performs the following processing using this.

On the basis of all the arc contours having a size sufficient to take the number of sampling points with respect to the arc contour stored in the arc contour parameter 9a (step 54),
The radius and center coordinates are obtained by the least squares method (step 55). Then, a search is made as to whether circular measurement is possible by the following method. (1) Using the center coordinates calculated by the least-squares method in the pattern of FIG. 4A, the angle θ formed by two vectors of the starting point P ₉ → center point → end point P ₁₀ of the circular arc contour is obtained, and
The ratio of 0 degrees is calculated. If the angle is smaller than the identification relative allowable angle, there is a possibility of erroneous recognition. (2) In the case of the arc contour lines i and k in FIG. 4A, it is determined whether or not a plurality of the same circles exist. The respective differences between the respective radii and the respective center coordinates are compared (step 56 in FIG. 3). If the differences are within the discrimination relative allowable range, it is considered that they have arisen from the same circle, and a plurality of arc contour lines are integrated. (Step 57) The measurement is again performed by the least squares method, and the content of the above (1) is performed (Step 58). (3) As shown in FIG. 7 (A), a state in which a circular protruding object E overlaps with the circle A in FIG. 4 (A) will be described. When the calculated center coordinate O of the circle is inside the circle A, there is a possibility of erroneous recognition, so that it is excluded from the measurement target. (4) Referring to FIG. 7B, the case where the protrusion C or the like is in contact with the circle A and the inflection point is difficult to detect as shown in FIG. 4A will be described with reference to FIG. When a circle (circle including a broken line) larger than the circle A on the screen is calculated by the arc k) in the vicinity of the contact, first, the center point S and the circle A of the calculated circle (broken line) are calculated.
(Solid line) arc sampling points (Q ₁ , Q ₂ , ..., Q
_n ) and Q ₁ S = r ₁ , Q ₂ S = r ₂ , Q _n S =
It is defined as r _n, the radius of operation circle with R. At this _{time, ((R-r 1)} 2 + (R-r 2) 2 + ... + (R-r n) 2) / n
> Compare the calculated circle with the actual circle on the screen using a discriminant that gives an identification relative allowable value (where n is the number of sampling points). If this discriminant is established, it is not measured.

As described above, the circle is verified, and if it can be measured, the size of the circle is calculated (steps 59 to 61 in FIG. 3).

On the other hand, based on all the side contour lines having a size sufficient to take the number of sampling points with respect to the side contour line stored in the side contour line parameter 9b, the circular and square extraction unit 10 performs the following method. A quadrangle is extracted from the irregular shape to search whether it can be measured. When a polygon in which the irregular shape D overlaps the quadrangle B as shown in FIG.
Under the following conditions, it is a measurement target. Conversely, if the following conditions are not satisfied, the measurement is not performed. (1) internal angle of the vector V ₂ of a vector V ₁ and the following side b in side a must be perpendicular. (FIG. 8 (A)) (2) V 1 and the vector V ₃ are parallel orientation with sides g it must be reversed. (FIG. 8B) (3) There must be an overlap between V ₁ and V ₃ . (FIG. 8 (C)) As described above, if the measurement can be performed after the verification of the rectangle, numbering is performed on each side to calculate the distance (the size of the rectangle) between two facing sides. When each of the circle and the rectangle is measured, the measurement result is registered (step 6).
2).

As described above, according to the first embodiment of the present invention, even if a plurality of types of circles (including a shape close to a circle), squares overlap or irregular shapes overlap, the circles and squares are erroneously determined. Extraction can be performed quickly without any measurement. Therefore, for example, in the case of measuring and adjusting the shape of tableware in a dining room, it is possible to instantaneously judge and settle even an image having irregular projections on circular tableware or square tableware.

FIG. 9 is a block diagram of a second embodiment of the present invention, in which parts having the same functions as those in FIG. 2 are assigned the same reference numerals and explanations thereof will be omitted. In FIG. 9, reference numeral 15 denotes a shape pattern formed by overlapping polygons. Reference numeral 16 denotes a measurement processing device which performs processing according to the flowchart shown in FIG.

As shown in FIG. 10A, a contour point is detected by a contour tracing method in a counterclockwise direction with respect to a shape pattern in which a quadrangle K and a polygon L overlap (step 50 in FIG. 3). Following the line, the contour point coordinates (x, y) are extracted and vectorized (step 51) and stored in the contour point memory 17. FIG. 10B shows the directions of the angles formed by the vectors at the inflection points P ₁ and P ₄ . A vector connects adjacent contour points counterclockwise. The angle between the vectors is
The vectors V _m and V _{n of interest} and the previous vector V
_The ± direction is defined and calculated using _m−2 and V _n−2 as shown in FIG. 10B (step 52 in FIG. 3). When the angle formed by the vector suddenly becomes negative, the inflection point P ₁ is obtained, and when the angle formed by the vector suddenly becomes positive, the inflection point P ₄ is obtained.

However, FIG. 10 shows an embodiment, and the tracing of the contour can be performed in the clockwise or counterclockwise direction, and the sign of the angle formed by the vector can also be defined. .

In FIG. 10A, the inflection points are P ₁ ,
P _2, ..., P _4, next to ... P _8, the sides (the part edge contour) a between the inflection point, b, ..., e, ... , coordinates sides contour parameters of this portion side contour as h 18 is stored. Next, the quadrangle extraction unit 19 performs the following processing using this.

On the basis of all the edge contours having a size sufficient to take the number of sampling points with respect to the edge contour stored in the edge contour parameter 18 (step 54), a polygon is converted into a rectangle by the following method. Extract and verify if it can be measured. (A) The representative vectors are defined as V _c and V _g, as shown in FIG. 11, for the vectors of the sides c and g in FIG. If V _c and V _g are generated from the same side, 1
Integrating the vector V _cg of the present (step 70,7
1).

For example, the object is measured under the following conditions. Conversely, if the following conditions are not satisfied, the measurement is not performed.

[0027] (1) the direction of the vector V _c and V _g must be oriented in the same direction.

[0028] (FIG. 12 (A)) (2) the distance d of the vector V _c and V _g must be less than the identified relative tolerance. (FIG. 12 (B)) 0 ≦ d < identify the relative tolerance (3) and the direction of the vector V _d direction and ending after start previous vector V _b of the vector V _c are the same, and the vector V _g It must be the same starting point and the direction of the previous vector V _f and the direction of the end point after the vector V _h of
. (FIG. 12 (C)) (4) vector V _c and V _g of the starting point previous vector V _b, V
The directions of _f must be opposite to each other. (FIG. 12 (C)) Here, as shown in FIG. 13, for example, there is a positional relationship in which another rectangle T overlaps between two identical rectangles R and S, and the above (1) and (2) If they are satisfied, V _C and V _g may be integrated into one line. In this case, where the two rectangles R and S must be separately identified, the two rectangles R and S are measured as one rectangle. To prevent this, the vector V _c, starting earlier the vector V _b relative to V _g, V _f,
FIG. 12 shows vectors V _d and V _h after the end point.
Verification is performed as shown in FIG. That is, V _b ,
_If there are V _d and V _{f on} the opposite side of V _h , they are not measured. (B) The representative vectors for the sides a, e, and g in FIG. 10A are defined as V _a , V _e , and V _{g in} FIG. 11, and the state of the two opposing sides is verified (FIG. Steps 72-7
4). For example, under the following conditions, the object is measured, and conversely, under the following conditions, the object is not measured.

(1) Vectors V _a and V _g must be parallel and opposite in direction. (FIG. 14 (A)) (2) Vectors V _a and V _g must have an overlapping portion. (FIG. 14 (B)) (3) When there are two or more matching side vectors V _e and V _g with respect to one target side vector V _a ,
Distance shall mutually directed vector V _g the nearest side (when the distance d _1, d ₂ is present, is regarded to have protruding object, taking whichever is closer). (FIG. 14 (C)) (c) The area between the vectors V _b and V _h of the two opposing sides as shown in FIGS. 11 and 13 is excluded from the measurement target under the following conditions.

(1) In FIG. 13, two vectors V _c and V _{e which} are perpendicular to the two vectors V _b and V _h and have the same direction are present. (FIG. 15 (A)) (2) A case where there is no overlapping part between two vectors V _a and V _cg perpendicular to the two side vectors V _b and V _h in FIG. 11 (FIG. 15 (B)) If either one of the two conditions occurs, it is excluded from the measurement target (step 75 in FIG. 3).

As described above, if it is measurable after the search for the square, numbering is performed on each side, and the distance (size of the square) between the two sides facing each other is calculated (step 76). If measurement is not possible, it is regarded as another polygon and measurement is not performed. After measuring the square, the measurement result is registered (step 62).

As described above, according to the second embodiment of the present invention, even if a plurality of kinds of polygons are superimposed, it is possible to quickly measure a quadrangle from there without error. Therefore, even in the case of measuring the shape of the tableware in the dining room and adjusting the price, even for an image with a polygonal protrusion on a square tableware,
Judgment and settlement can be done instantly.

[0033]

According to the present invention, it is possible to accurately and quickly extract a circle and a square even when a plurality of overlapping circles or squares or an overlap of irregular shapes is detected. On the other hand, even when a plurality of polygons are overlapped. A rectangle can be accurately and quickly extracted. Therefore, for example, when the present invention is applied to a system for measuring and adjusting the shape of tableware in a cafeteria, the shape of the tableware can be accurately determined even if the tableware protrudes.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating the principle of the present invention.

FIG. 2 is a block diagram of a first embodiment of the present invention.

FIG. 3 is an operation flowchart of the present invention.

FIG. 4 is a pattern in the first embodiment.

FIG. 5 is a diagram illustrating equalization of a partial contour line.

FIG. 6 is a diagram illustrating a state in which a partial contour is stored as a parameter.

FIG. 7 is a diagram for explaining identification of a pattern having an overhang;

FIG. 8 is a diagram illustrating conditions when an irregular shape overlaps a square.

FIG. 9 is a block diagram of a second embodiment of the present invention.

FIG. 10 is a pattern in the second embodiment.

FIG. 11 is a vector (indefinite form) in the second embodiment.

FIG. 12 is a diagram illustrating conditions for integrating as one vector.

FIG. 13 is a vector (square) in the second embodiment.

FIG. 14 is a diagram illustrating conditions of two upper and lower sides facing each other.

FIG. 15 is a diagram for describing conditions of two left and right sides facing each other.

FIG. 16 is a pattern in which circles overlap.

FIG. 17 is a diagram illustrating sampling of a pattern in which circles overlap;

FIG. 18 is a diagram for explaining a polygon identification method in a conventional example.

FIG. 19 is a numbering diagram of the pattern in FIG. 18;

FIG. 20 is a pattern in which polygons are overlapped.

FIG. 21 is a diagram illustrating a polygon identification method in a conventional example.

FIG. 22 is a numbering diagram of the pattern shown in FIG. 21;

FIG. 23 is a diagram illustrating a problem of the conventional example.

[Explanation of symbols]

 Reference Signs List 1,15 Shape pattern 3 Image input device 6 Image memory 7,16 Measurement processing device 8,17 Contour point memory 9a Circular contour parameter 9b, 18 Side contour parameter 10 Circular and square extraction unit 19 Square extraction unit A Circular B , K, R, S, T Rectangle C, D, L Indefinite (protrude)

────────────────────────────────────────────────── ─── Continuing on the front page (72) Inventor Tsutomu Wakimizu 17-58 Higashi-Kasuga-cho, Oita City, Oita Prefecture Fujitsu Oita Software Laboratory Co., Ltd. (56) References JP-A-4-160575 (JP, A) JP-A-64-70890 (JP, A) JP-A-1-161487 (JP, A) JP-A-60-204086 (JP, A) A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G06T 1/00 G06T 5/00 G06T 7/00 G06T 7/60

Claims (4)

(57) [Claims] 1. A single pattern of a circle (including a shape close to a circle) or a square, or a pattern in which a plurality of circles and rectangles are respectively superimposed on the same screen, or an unknown irregular shape pattern is included in the pattern. When superimposed, contour points are extracted, sample points are taken at appropriate intervals, vectorized, an inflection point is detected from a change in the direction of the angle between the two vectors, and the contour line between the inflection points is detected. The partial contour is equally divided, and the angle between the vector of interest and the vector preceding the vector is sequentially calculated. If the average value of the sequentially calculated angles is equal to or greater than a set threshold, the partial arc contour In the following, a pattern is used to separate and identify the partial side contour lines and to identify what kind of circles and squares are generating patterns on the screen by using the lengths of the two partial contour lines. In the case of a circular shape, in the case of a circle, the angle formed by the partial arc contours is equal to or greater than a threshold value set for all the arcs (condition 100), and the partial arc contours are separated and identified by the inflection point . The radius and the center of the calculation circle based on the center point of the calculation circle calculated from the partial arc contour line having no inflection point where the center coordinates of the circle do not exist inside the separately identified circle (condition 101). The difference between the point and the radius and center point of the arc contour line on the actual screen is equal to or smaller than the threshold value (condition 102).
102) are excluded as indefinite if not satisfied,
Next, in the case of a quadrangle, the interior angle at the inflection point (the angle with the vector of the next side) is a right angle (condition 103), and the vectors at the two opposing partial side contour lines are parallel and the directions are Be opposite (condition 104) and have an overlap between the two vectors (condition 10
According to 5), any rectangle is identified, and the rectangle identification condition (103
To 105), a pattern identification method characterized by excluding all the irregular shapes if they do not satisfy the condition. 2. When a pattern on which a plurality of polygons are superimposed is present on a screen, contour points are extracted, sample points are taken at appropriate intervals, vectorized, and an inflection point is obtained from a change in the direction of the angle formed by the two vectors. Is detected and separated into partial side contours, which are sides between the inflection points, and the type of quadrangle which generates a pattern on the screen by using the partial contours is identified. If there are a plurality of partial side contours generated from the same side, a vector of two opposing partial side contours including the integrated partial side contours Are parallel and opposite in direction (condition 110), have an overlapping portion between two partial edge contour vectors (condition 111), and face a plurality of target partial edge contour vectors in plural. Partial ring When the vector of the contour line exists, by selecting the vector of the nearest partial side contour line (condition 112), two opposing vectors are extracted, a rectangle is identified from the polygon, and the rectangle identification condition (110) is selected. To 112), the pattern identification method is characterized in that all the patterns are excluded as indefinite if they do not satisfy the condition. 3. When there are a plurality of partial side contours generated from the same side, in a method of integrating these into one straight line, if two vectors are generated on the same side, The directions of the two vectors are in the same direction, and the distance between the two vectors in the direction perpendicular to this direction is smaller than the discrimination relative tolerance, and the direction of the vector before the start point and the direction after the end point in each vector. 3. The two vectors are integrated into one vector because the directions of the vectors are the same and the directions of the vectors before the start point of the two vectors are opposite to each other. Pattern identification method. 4. In the method for extracting two opposite vectors, if there is a vector orthogonal to the extracted two opposite vectors, the two different vectors having the same direction vector at right angles to the two opposite vectors. If there are two vectors, and if any of the two conditions for the case where there is no overlapping part between the two opposing vectors and the vector perpendicular to the two vectors are satisfied, the rectangle is excluded from the measurement target,
Also, if there are two different vectors having a vector that is perpendicular and opposite to the two opposing vectors, and if there is an overlap between the two opposing vectors and the vector that is orthogonal to the two opposing vectors, 4. The pattern identification method according to claim 3, wherein when one of the conditions is satisfied, a square is to be measured.