JPH0528274A - Automatic plane point pattern-matching device and matching method - Google Patents

Abstract

PURPOSE: To make it possible to execute more precise matching of such patterns which vary in the number of repeating points in two patterns and with which the prediction of the probability of the number of these points is not possible by using a contrast probability calculator, contrast point selector and similarity calculator. CONSTITUTION: The respective patterns in, for example, three kinds of first to third plane point patterns are composed of 8 pieces of points 1 to 8. The contrast probability calculator is first used and the contrastable pair is selected by an approximate matching method in order to start the matching of the first pattern and the second pattern. A fuzzy relaxation method is used for calculating the probability that the points 1 to 8 in the first pattern contrast with all the points 1 to 8 of the second pattern is used in the contrast point selector. The contrast probability, average contrast probability, the average distance between the contrast points and radix numeration are used in order to determine whether the two patterns are based on the same origin or not in the similarity calculator.

Description

Detailed Description of the Invention

[0001]

FIELD OF THE INVENTION The present invention relates generally to apparatus and methods for matching planar dot patterns, and more particularly to matching digitized computer images such as fingerprints, linguistic characters, and planar images.

[0002]

BACKGROUND OF THE INVENTION Computer vision has become one of the hot topics in the development of fifth generation computers. Computer vision consists of two areas: image processing and pattern recognition. Processing the pattern reveals individual characteristics for recognition. For example, the minutiae of the fingerprint are obtained with the processing operation, and if the fingerprint is specific, these minutiae are used for the appraisal.

Computer reading or scanning of linguistic characters can also be used to recognize the characters by the dot pattern recognition method. Handwritten Chinese characters are shown on the left side of FIG. After processing, the characteristic points obtained by the computer are shown in FIG.
Shown on the right side of. The left hand side of FIG. 2 shows the same handwritten Chinese character with different writing styles, and the right hand side of FIG. 2 shows the characteristic points of this Chinese character. The computer recognizes this character by comparing the point patterns on the right side of FIG. 1 and on the right side of FIG. 2 to determine if these patterns are feature point patterns obtained from the same Chinese character. Knowing the kanji, it can be seen that the left side of FIG. 1 and the left side of FIG. 2 are the same. However, the computer makes the decision only by the special procedure described below.

Generally, there are two types of operations in the comparison procedure: matching and registration. The matching procedure is to find a one-to-one correspondence between the points of the first plane point pattern and the points of the second plane point pattern. That is, there is one point in the first plane point pattern and only one point in the second plane point pattern is involved.

The registration procedure is performed by translating, rotating and / or enlarging / reducing the first plane point pattern on the Euclidean plane to obtain the second plane point pattern (reference plane point pattern).
To obtain exactly the same pattern as, and to investigate the similarity between them more precisely.

Image processing for handwritten characters and fingerprints and pattern recognition have the following two major differences. (1) The number of feature points is different between the two plane point patterns, and (2) the probability of feature points appearing in both patterns cannot be predicted. For example, the feature points on the right side of FIG. 2 should theoretically be the same, but they are different in practice. In the figure, the p ¹ and q ¹ portions surrounded by a frame are completely different from the p ² and q ² portions. Not only the number of points is different, but the distribution of points is also different, which makes prediction difficult.

Taking a fingerprint as another example, the fingerprint identifies a person and cannot be changed for a lifetime.
Fingerprint recognition using computer image processing methods to determine whether two fingerprints are unique is a goal scientists have long sought. Since the fingertips are soft and supple, the fingerprints usually collapse when the fingertips are pressed against the glass surface. Therefore, it is impossible to show the fingerprint completely on the pattern to be processed. Statistically, the probability that a fingerprint feature will reproduce two patterns is only 60.
%.

Furthermore, the pressure exerted by the distribution state of the fingertip also affects the distribution of the characteristic points. Generally, the pattern rotates or translates depending on the direction of pressure. Therefore, it becomes more difficult to compare fingerprint patterns.

[0009]

OBJECTS OF THE INVENTION Therefore, an object of the present invention is to provide an automatic matching for patterns that differ in the number of repeating points in both patterns and / or in which the probability of the number of these points is unpredictable. Plane dot pattern
It is to provide a matching device.

Another object of the present invention is to provide an automatic plane point pattern matching device for more precisely matching two patterns even if the patterns are rotated, translated and / or scaled. is there.

Another object of the present invention is to provide a rapid automatic plane point pattern matching device.

Still another object of the present invention is to provide a method for providing the above pattern matching function.

[0013]

In order to achieve these objects, the present invention uses an automatic probability point calculation device that performs matching between two optimum contrast patterns using a contrast probability calculation device, a contrast point selection device, and a similarity calculation device. A matching device is provided.

The contrast probability calculation apparatus of the present invention uses the "fuzzy relaxation method" to calculate the probability that a point in the first pattern is contrasted with all the points in the second pattern. The contrast point selection device uses the "continuous forward selection method" to select a pair of appropriate contrast points. In the similarity calculator, "contrast probability", "mean contrast probability", "average distance between contrast points" and " The "scale factor" is used.

Further, in order to improve the processing efficiency, in a specific example described later, a "moving device" is provided to reduce the processing time by removing the relationship between the uncomparable points. In the moving device according to one embodiment of the present invention, the "rough matching method" is used to perform the movement.

Furthermore, the present invention provides a registration device that uses the "minimum mean square error" to rotate, translate and / or scale the planar point pattern to be matched to the most suitable state for comparison. .

[0017]

First, the fuzzy relaxation method, the continuous forward selection method, and the similarity calculation method will be briefly described.

Fuzzy relaxation method P = {p ₁ , p ₂ , ...
,, p _m }, consider two plane point patterns represented by Q = {q ₁ , q ₂ , ..., Q _n }. The initial contrast probability is (i = 1, 2, ..., Each pair of (p _i , q _j ).
m; j = 1, 2, ..., N). j
= 1, 2, ..., N, the point p _i of the set P contrasts only the point q _{j of the} set Q, so there is no difference.
It is clear that all pairs give the same initial contrast probability. Then, when (p _i , q _j ) is the correct contrast pair, the contrast probability is iteratively modified according to the probability of the other point p _h to correctly contrast the other point q _k in Q. S
^{_{(O) (p i, q}} j) pairs _(p i, _{q j)} represents the initial contrast probability, and represents the contrast probability of the pairs in _{^{S (r) (p i,}} q j) is the order r version Then, these relationships are as follows.

[0019]

[Equation 20]

The reason for performing the "maximum" operation is p _i and q
_This is because it is assumed that _j has a one-to-one correspondence. Symbol C
_ij (h, k) represents the probability measure of another pair (p _h , q _k ) when (p _i , q _j ) is the correct contrast pair, and C _ij (h, k) is Is defined as

[0021]

[Equation 21]

[0022]

[Equation 22]

Here, 1 is a symbol representing the distance between the points indicated by the subscript.

0 ≦ Δ ≦ 1, 0.5 ≦ C _ij (h, k) ≦
Since it is 1, the fitness measure simply decreases from 1.
By iteratively modifying this contrast probability, the correct contrast pair keeps a larger value while the inaccurate contrast pair approaches zero.

The result of the above calculation gives the possibility that the points belonging to the first pattern are correctly compared with the points belonging to the second pattern.

Approximate Matching Method In a coordinate system, if two points to be contrasted (they belong to different patterns) are separated from each other, they have no chance to be a correct contrasting pair, and the fuzzy relaxation method. After treatment by
It is clear that the contrast probability approaches zero. Therefore, if the contrast probability of an apparently inaccurate contrast pair becomes zero very early, and the fuzzy relaxation method only applies to nonzero probabilities, the mathematical operations are greatly reduced. This increases the overall efficiency of this method.

The present invention provides a coarse matching method to eliminate apparently inaccurate contrast pairs. As a general rule, in this rough matching method, the contrast probabilities of points having a distance from each other are set to zero. The criteria for determining whether a point is far from its contrast point are:

[0028]

[Equation 23]

Here, X and Y represent the x-coordinate and the y-coordinate, respectively. The value n depends on the origin of the associated pattern. For example, fingerprints and letters have different values for n.

Continuous Forward Selection Method Set of Two Points P and Q
The optimal contrast pair between is obtained by the fuzzy relaxation method in which the contrast probability is repeatedly modified several times as described above. The method of selecting the optimal contrast pair from the matrix composed of the probabilities of all pairs (each pair includes the points of the first pattern and the points of the second pattern) is not a unique method. However, conventional methods such as linear programming, primal analyses, out of quilter algorithms, etc. are not useful in this case. This is because the sampling environment and the sampling method influence the image to be used to some extent, so that even if the sample point patterns P and Q show the same arrangement, some difference occurs. The most noticeable difference is (a) different sets P
The number of points of both Q and Q, and (b) the probability of the same point appearing in both sets cannot be determined (some point appears in P, but Q
Does not appear in, and vice versa. That is.

In order to overcome this difficulty, a new method called "continuous forward selection method" has been developed. To use this method, first create the matrix [S _ij ]. Where i = 1,
2, ..., M, j = 1, 2, ..., N, m is the number of points in the pattern P, and n is the number of points in the pattern Q. The element S _{ij of} this matrix is q _j of p _i .
Is the contrast probability for. Then, the largest element K is selected from the matrix [S _ij ] and the set S is created using this K. Here, K = min (m, n). This method develops as follows.

(A) The maximum element S _max of S is placed in the matrix [S _ij ], and the associated pair (p _i , q
_j ) are correctly contrasted.

(B) The elements in the same row or the same column of (p _i , q _j ) are removed. That is, S _ij is removed from the matrix [S _ij ].

(C) S _max is removed from S.

(D) The steps (a) to (c) are repeated until S becomes empty.

It may be impossible to select all the correct contrast pairs in the above procedure, but they have the same origin,
It is still a good and perhaps optimal way to handle patterns where the number of points is different and the probability that the same point will appear in both patterns is unpredictable.

Similarity Calculation Method With respect to two patterns that are compared with each other, the number of points that can be correctly compared, the comparison probability of each pair, and the coordinates of these points are calculated by the notation method. Then, using the parameters thus obtained,
Calculate (1) pattern matching probability (X ₁ ), (2) average contrast probability (X ₂ ), (3) average distance between contrast pairs (X ₃ ), and (4) scale factor (X ₄ ). . These four coefficients are defined as follows.

(A) Matching probability (X ₁ ):

[0039]

[Equation 24]

(B) Mean contrast probability (X ₂ ):

[0041]

[Equation 25]

(C) Average distance between contrast pairs (X ₃ ):

[0043]

[Equation 26]

(D) Scale factor (X ₄ ): r ₃ = S · c
If os θ and r ₄ = S · sin θ,

[0045]

[Equation 27]

Is the magnification and X ₄ is

[0047]

[Equation 28]

Is defined by For two similar plane point patterns, S → 0, X ₄ → 1, else X
₄ > 1.

From the above definition, if the two sets P and Q are very similar, then X ₁ → 1, X ₂ → 1, X ₃ → 0 and X ₄ → 1, otherwise X ₁ → 0, X ₂ → 0, X
_It can be concluded that ₃ > 0 and X ₄ > 0. The score representing the similarity between P and Q can be obtained from X ₁ , X ₂ , X ₃ and X ₄ . That is,

[0050]

[Equation 29]

Here, C is a constant. For the purpose of statistically analyzing a large number of cases, as shown in Fig. 3,
A curve of points is obtained according to the above definition. Therefore,
It can be seen that if the score is T or higher, the pattern is generated from the arrangement of the same origin. If the score is T or less, no pattern is generated from the same origin. For example, a very large number of fingerprints (including both the same fingerprint and different fingerprints).
Is sampled and statistically analyzed to obtain a threshold value (T). This threshold can be used to determine if a fingerprint is the same as a particular fingerprint. It is clear that this arbitrary fingerprint is the same as the reference if the score obtained using the arbitrary fingerprint and the reference fingerprint is above the threshold value.

For those skilled in the art, these four coefficients X
It is very easy to give different weighting factors to ₁ , X ₂ , X ₃ and X ₄ to adjust the point values to suit the requirements of different purposes. Since the pattern is often translated, rotated, and / or magnified before sampling, the X ₃ value obtained from equation (6) is increased and the score thus obtained is not suitable. .

In order to avoid this kind of situation and to more accurately match the two patterns, the pattern registration should first be done before comparison with the given reference.

Registration of patterns using the minimum mean square error method will be described below.

Minimum Mean Square Error Registration When a contrast pair is chosen, the points in set Q are rotated, translated and / or expanded to minimize the average distance between the contrast points in set P and set Q. . Then, the similarity between P and Q can be calculated using the registered pattern. The minimum distance between them is calculated by the following minimum mean square error method.

Consider the transformation R. R is the point q of Q
Convert _mi (x _mi , y _mi ) (i = 1, 2, ..., K) into q ¹ _mi (x ¹ _mi , y ¹ _mi ) to reduce the following sum.

[0057]

[Equation 30]

The conversion of q _mi is as follows.

[0059]

[Equation 31]

Where r ₁ and r ₂ are the amounts of translation, r
₃ and r ₄ represent the amount of rotation. By using the above conversion, the pattern can be registered most appropriately to keep X ₃ low.

[0061]

SPECIFIC EXAMPLE FIG. 4 shows three types of plane point patterns A, B and C. Each pattern consists of 8 points, and their coordinates are also shown in FIG. Note that pattern B is obtained from pattern A by appropriate translation and scanning. Pattern C is clearly not the same as pattern A.

(A) Matching between pattern A and pattern B In order to start matching, first, a contrast probability calculator is used to select a pair that can be compared by the rough matching method. In this example, pairs having x- or y-coordinates different from each other by 30 (that is, n = 30) or more are excluded. The result is shown in FIG. The initial contrast probabilities of these contrastable pairs are all set to 1, while those excluded by the coarse matching method are set to zero.

Next, the contrast probability is corrected using the equation (1). The results of two corrections are shown in FIG. The maximum contrast probability 8 is selected by the contrast point selection device using the continuous forward selection method. These are listed in order as follows.

K = 8S = {0.8858, 0.885
2, 0.8691, 0.8640, 0.8559, 0.
8260, 0.8251, 0.7947}

8 pairs of points related to the above-mentioned contrast probability 8 are obtained, and these are shown in FIG. In the figure, 1 is a comparison
0 represents an unpaired pair.

With the similarity calculator, the coefficients X ₁ , X ₂ ,
X ₃ and X ₄ are calculated. These are as follows.

X ₁ = 1 From Expression (4), X ₂ =
0.8502 From Expression (5) X ₃ = 6.4780 From Expression (6) From X ₄ = 1 From Expression (7)

In this example, when the constant C is 100,000, the score obtained by the equation (8) is 13124.42.
Is. The threshold used here is the number 13124.
If it is 42 or less, that is, T <score-13124.42, the patterns are similar, and if it is the above numerical value or more, they are not similar.

(B) Matching between pattern A and pattern C By the same procedure as used in the above example, the initial contrast probability and the modified contrast probability are obtained and shown in FIGS. 8 and 9, respectively. The results are as follows.

K = 8S = {0.8402, 0.840
2, 0.8402, 0.8402, 0.8402, 0.
8402, 0.8402, 0.8402}

From the above result, the matching relationship between the two patterns can be obtained. These are shown in FIG. p ₃ and q
It can be seen that ₂ , p ₄ and q ₄ , p ₅ and q _6, and p ₅ and q ₈ are respectively compared. From the calculation of similarity, X
_{1 = 0.25, X 2: 0.8402} , X 3: 15.79
10 and X ₄ = 1 are obtained, thus C = 100.0
If it is 00, the score is 133.18. Since this score is much smaller than the threshold, the conclusion is that patterns A and C are not similar.

[0072]

It is clear that the apparatus and method of the present invention can precisely match two patterns even if the pattern to be matched is broken. Such matching is not possible with conventional equipment. Further, even under the condition that the number of points in the pattern is different, matching of similar patterns is effectively performed in the present invention.

Described herein is a method and apparatus for automatically matching planar dot patterns. Various details of the invention can be changed without departing from the spirit of the invention. Furthermore, the above description of the preferred embodiments of the present invention has been presented for purposes of illustration and is not intended to limit the invention. The invention is limited only by the claims.

[Brief description of drawings]

FIG. 1 shows a handwritten Chinese character input from an image scanner on the left side and a distribution of feature points related to the handwritten character shown on the left side on the right side.

2 shows another handwritten typeface of the character shown in FIG. 1 on the left side, and the right side shows the distribution of feature points related to the handwritten character shown on the left side.

FIG. 3 shows a similarity distribution between patterns after being processed by the similarity calculation device of the present invention.

FIG. 4 shows three types of plane point patterns used as examples for explaining the present invention.

FIG. 5 shows an initial contrast probability when matching the patterns A and B of FIG. 4 according to the rough matching method.

6 is a modified contrast probability obtained from the data of FIG. 5 by the fuzzy relaxation method.

FIG. 7 shows the results of the data described in FIG. 6 processed by the continuous forward method.

FIG. 8 shows initial contrast probabilities when matching the patterns of FIG. 4 according to the rough matching method.

FIG. 9 shows a modified contrast probability obtained from the data of FIG. 8 by the fuzzy relaxation method.

FIG. 10 shows the results of the data described in FIG. 9 processed by the continuous forward selection method, respectively.

Claims (22)

[Claims] 1. A contrast probability calculation device that uses a relaxation method to calculate a probability of a point of a first plane point pattern and compare it with a probability of a point of a second plane point pattern; a point of the first pattern; A contrast point selection device that uses a continuous forward selection method to select an optimal contrast pair for the points of the second pattern, and the contrast probability calculation device that calculates a similarity value between the first and second patterns. The comparison point selection device and / or a similarity value calculation device that uses a result obtained by the coordinates of the points of the first and second patterns, wherein the matching device is configured to perform the comparison when the similarity value is a predetermined value or more. An automatic plane point pattern matching device, characterized in that the first and second patterns are determined to be similar. 2. The apparatus according to claim 1, wherein the contrast probability calculation device calculates a preset contrast probability between a point p _i of the first pattern and a point q _j of the second pattern as the other of the first pattern. , According to the probability for the point p _h of the second pattern to form a correct contrast pair with the other points q _k of the second pattern when the points p _i and q _j form a correct contrast pair. apparatus. 3. The apparatus according to claim 2, wherein the contrast probability calculation device repeatedly corrects the probability for the point p _i of the first pattern, and compares the probability with the point q _j of the second pattern according to the following equation. Automatic plane point pattern matching device: [Equation 1] Where S ^(r) (p _i , q _j ) is the probability in the rth version, m is the total number of points in the first pattern, C
_ij (h, k) is the probability of another point p _h of the first pattern to contrast another point q _k of the second pattern when the points p _i and q _j form a correct contrast pair. , S
^(O) (p _i , q _j ) is the initial contrast probability of the points p _i and q _j . 4. The device according to claim 3, wherein C
An automatic plane point pattern matching device in which _ij (h, k) is calculated by the following equation: And [Equation 3] Where 1 represents the distance between the points shown as subscripts in the above equation. 5. The apparatus according to claim 1, wherein the contrast probability calculating device includes an initial contrast probability calculating device, and the initial contrast probability is zero when a distance between paired points is a predetermined value or more. Automatic plane point pattern matching device to be set. 6. The apparatus of claim 5, wherein the distance between the pair of points is the distance along the X axis of the coordinate system. 7. The apparatus of claim 5, wherein the distance between the pair of points is the distance along the Y axis of the coordinate system. 8. The apparatus according to claim 1, wherein the contrast point selecting device includes contrast probability arranging means and contrast point searching means, and the contrast probability arranging means calculates the contrast calculated by the contrast probability calculating device. Probabilities are arranged in the order of maximum and minimum, and the same number of maximum probabilities is selected as the minimum number among the number of points of the first and second patterns, and the contrast point search means selects the maximum probability of the several types. An automatic plane point pattern matching device that finds a point related to any one of them as a correct contrast and an abandoned point that has already been contrasted. 9. The automatic plane point pattern matching device according to claim 1, wherein the similarity calculation device calculates the similarity between the first and second patterns using the following function: Contrast probability (X ₁ ) [Equation 4] Here, P represents the first pattern and Q represents the second pattern: (b) Average contrast probability (X ₂ ): Here, k is the number of contrast pairs, and S ^(r) (p _i , q _mi ) is the probability of p _i that contrasts the contrast point q _mi : (c)
The average distance (X ₃ ) between the points of the contrast pair: Here, k is the number of contrast pairs, and (X _i , Y _i ) and (x ¹ _mi , y ¹ _mi ) represent coordinates of p _i and q _mi , respectively; (d) scale factor (X ₄ ). [Equation 7] Here, S is a scale score. 10. The apparatus according to claim 9, wherein the similarity calculation device calculates the similarity between the first and second patterns using the following equation: ] Where C is a constant and the score represents the similarity between the first and second patterns. 11. At least a contrast probability calculating means for calculating a probability for one point of the first pattern and comparing it with an arbitrary point of the second pattern, and selecting an optimum contrast pair from the points of the first and second patterns. The comparison point selecting means and the similarity calculating means for calculating the similarity (value) between the first and second patterns, and the pattern is rotated (moved) so as to be easily compared.
A first plane point pattern composed of a plurality of points and a second plane point pattern composed of a plurality of points, including a matching means composed of registration means used for parallel movement and / or scaling. In the automatic plane point pattern matching device for matching, the matching means calculates the similarity of the first and second patterns using the adjusted pattern, and the similarity between the two is equal to or more than a predetermined value. If so, the automatic plane point pattern matching apparatus is characterized in that the two patterns are determined to be the same. 12. The apparatus according to claim 11, wherein said registration means corrects the pattern according to the following equation: Here, (x _mi , y _mi ) is (x ¹ _mi , y ¹ _mi )
, The original coordinates of (x ¹ _mi , y ¹ _mi ) represent the modified coordinates of said point, R is the transformation, r ₁ and r ₂ are the amounts of transformation, and r ₃ and r ₄ are the amounts of rotation. 13. A method for matching a first plane point pattern composed of a plurality of points and a second plane point pattern composed of a plurality of points, including at least the following steps. Automatic plane point pattern matching method: calculating a probability for one point in the first pattern, and calculating a contrast probability using a fuzzy relaxation method for comparing arbitrary points in the second pattern; And selecting a contrast point using a continuous forward selection method to select an optimum contrast pair from the points of the second and second patterns; the first and second patterns for calculating the similarity between the first and second patterns. Calculating similarity using the contrast probabilities and / or coordinates of the points of the two patterns; if the similarity is greater than or equal to a predetermined value, it is determined that the first and second patterns match. Decision process. 14. The method of claim 13, wherein the step of calculating the contrast probability comprises the following steps: an automatic plane point pattern matching method: an initial contrast probability between the first pattern and the second pattern is calculated. Setting according to a predetermined value to set the initial contrast probability; and the following points p _i
And q _j are the correct contrast pair, the point q _j of the second pattern is contrasted according to the probabilities for the other points p _h of the first pattern and the other points q _k of the second pattern. Modifying the probability of the points p _i of the first pattern to modify the contrast probability. 15. The method according to claim 14, wherein the step of modifying the contrast probability includes the step of modifying the contrast probability of the point p _i and the point q _{j according} to the following equation. : [Equation 10] Where S ^(r) (p _i , q _j ) is the probability in the r-th version, m is the total number of points in the first pattern, and C _ij (h, k) is the point p. Probability of another point p _h of the first pattern to contrast another point q _k of the second pattern when _i and q _j form a correct contrast pair, and S ^(o) (p _i , q _j ) is the initial contrast probability of the points p _i and q _j . 16. The method according to claim 15, wherein the step of correcting the contrast probability includes the step of calculating C _ij (h, k) according to the following equation: And [Equation 12] Where 1 represents the distance between the points shown as subscripts in the above equation. 17. The method according to claim 14, wherein the step of setting an initial contrast probability is the point p of the first pattern.
_i (x _pi , y _pi ) and the point q of the second pattern
_An automatic plane point pattern matching method including the step of setting the initial contrast probability of _j (x _qj , y _qj ) to zero under the following conditions: Here, n is a constant. 18. The method according to claim 13, wherein the step of selecting contrast points includes the following steps: (1) From the contrast probability given in the step of calculating the contrast probability. Extracting a probability group;
The group comprises several probabilities, which are the largest of all the contrast probabilities, the number being the smallest of the number of points in the first and second patterns. (2) a step of finding a point related to the maximum probability from the probability group extracted in the step (1); (3) a point related to the maximum probability from the group, and a related point from the first and second patterns Removing; and (4) repeating steps (1) to (3) until none of the groups are included. 19. The method of claim 13, wherein four coefficients are defined as follows: (1) contrast probability (X ₁ ), mean contrast probability (X ₂ ), (3) contrast pair points. An automatic plane point pattern matching method which performs the step of calculating the similarity according to the average distance (X ₃ ) and (4) scale factor (X ₄ ) between: [Equation 15] Here, k is the number of contrast pairs, and S ^(r) (p _i , q _mi ) is the p _i probability with which the contrast points q _mi are contrasted; Where k is the number of contrast pairs, and (x _i , y _i ) and (x ¹ _mi , y ¹ _mi ) represent the coordinates of p _i and q _mi , respectively; Here, S is the enlargement ratio. 20. The automatic plane point pattern matching method according to claim 19, wherein the step of calculating the similarity includes the step of calculating the number of similar points according to the following equation: Here, C is a constant. 21. The method of claim 13, comprising the steps of using rotation, translation and / or scaling to modify the pattern to best fit the match. 22. The method according to claim 21, wherein the registration step transforms the coordinates (x _mi , y _mi ) of the points into other coordinates (x ¹ _mi , y ¹ _mi ) according to the following formula.・ Matching method: Where R is the deformation, r ₁ and r ₂ are the amounts of translation, and r ₃ and r ₄ are the amounts of rotation.