JPS62236090A - Pattern collating system - Google Patents

Abstract

PURPOSE:To determine correctly the coordinate matching quantity of a searching pattern and a file pattern even when the ratio including a pseudo pair comes to be large in a pair characteristic point candidate by determining the assembling, in which the number of the candidates of a pair characteristic point after applying the coordinate conversion out of the candidate of a coordinate matching quantity comes to be maximum, as the coordinate matching quantity. CONSTITUTION:A matching quantity candidate extracting circuit 7 extracts the address value as the candidate of a coordinate matching quantity in the sequence from the high memory area of the data value of a deviation memory 5. These values are sent from a matching quantity candidate memory 8 to a coordinate converting circuit 2, based upon this, the coordinate matching quantity is made into alpha=DELTAXi, beta=DELTAYi and gamma=DELTATHETAk and positions (x) and (y) and a direction theta of respective characteristic points of a searching pattern are coordinate-converted. A pair characteristic point candidate extracting circuit extracts the pair characteristic point candidate of this and a file pattern and obtains the number of the candidate. The number of the pair characteristic candidate is obtained to all assemblings of the coordinate matching quantities alpha=DELTAXi, beta=DELTAYi and gamma=DELTATHETAk, and the assembling, in which the number of the candidate comes to be maximum, is determined to the coordinate matching quantity of a searching pattern and a file pattern.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a pattern matching method that automatically verifies the identity of striped patterns represented by fingerprints using coded pattern features.

[Conventional technology]

In an automatic matching method for striped patterns such as fingerprints, the end points where the striped pattern breaks and the branch points where the striped pattern branches are taken as feature points of the pattern, and Rx
A quantity (pattern feature) representing the characteristics of each feature point including ey and direction 0 is coded and stored, and using this coded pattern feature, □The search pattern to be matched is registered in advance. Compare and match file pattern.

However, since the coordinate system of the search pattern and the coordinate system of the file pattern usually do not match, it is necessary to make their coordinate systems almost the same before comparing and matching them.

Figure 2 is shown, for example, in Japanese Unexamined Patent Publication No. 59-778.
FIG. 2 is a block diagram illustrating a conventional pattern matching method in which coordinate transformation as described above is performed. In the figure, 1 is a feature point memory, 2 is a coordinate conversion circuit connected to this feature point memory 1, 3 is a paired feature point candidate extraction circuit connected to this coordinate conversion circuit 2, and 4 is this paired feature point candidate extraction circuit. A paired feature point candidate memory connected to the circuit 3; 5 a deviation memory branched from the paired feature point candidate extraction circuit 3; 6 a matching amount connected to this deviation memory 5 and sending an output to the coordinate conversion circuit 2; This is an extraction circuit.

Next, the operation will be explained. Each feature point of the search pattern and file pattern is coded and stored in the feature point memory 1 as a pattern feature including its position XvV and direction 0. The coordinate conversion circuit 2 converts the pattern feature position IP2xty, the coordinate matching amounts α, β, corresponding to the direction 0,
This is converted by γ, and first the coordinate matching amounts α and β are calculated.

All of γ are set to '0'' and the pattern features stored in the feature point memory 1 are sent to the feature point candidate extraction circuit 3 without conversion. The points are sequentially compared, and feature points whose pattern feature differences are within a predetermined threshold are extracted as candidates for paired feature points.

It is stored in the paired feature point candidate memory 4, and weights are added to the data in the storage area of the deviation memory 5 that are associated with the positional differences ΔX, Δy and direction difference Δ0 of the paired feature point candidates.

However, all initial values of data in each storage area are It OI
Let it be t.

Here, if you repeat the above operation for all paired minutiae candidates extracted from the search pattern and file pattern,
If both patterns are of the same fingerprint, they are concentrated in the storage area of the deviation memory 5 corresponding to the coordinate matching amounts α, β, and γ of both patterns, and if they are not the same fingerprint, they are scattered in the storage area of the deviation memory 5. I don't think it's possible to concentrate. Therefore, the matching amount extraction circuit 6 extracts the address values ΔX, ΔY, and Δθ of the storage area having the maximum data on the deviation memory 5, and uses these as the coordinate matching amounts for both patterns, and feeds them back to the coordinate conversion circuit 2. . The position XpV and direction 0 of each feature point of the search pattern are the coordinate matching amounts α=ΔX, β=ΔY in the coordinate conversion circuit 2.

After coordinate transformation using γ=Δe, the search pattern and file pattern are compared and matched.

[Problem that the invention seeks to solve]

The conventional pattern matching method is configured as described above, and extracts the amount of coordinate matching based on the address value of the storage area where the data is concentrated the most on the deviation memory 5.
If there is a large proportion of paired minutiae candidates that contain false pairs that cannot actually be paired, the storage area where the data is most concentrated in the deviation memory does not necessarily match the amount of coordinate consistency. In that case, there is a problem that appropriate locus wA conversion is not performed.

This invention was made to solve the above problems, and even if the proportion of false pairs among paired minutiae candidates increases, the amount of coordinate matching between the search pattern and the file pattern can be accurately determined. can be determined.

The purpose of this paper is to obtain a pattern matching method that can perform appropriate coordinate transformation.

[Means for solving problems]

The pattern matching method according to the present invention includes a matching amount candidate extraction circuit that extracts several coordinate matching amount candidates from a deviation memory, and a matching amount candidate memory that stores the values of the extracted coordinate matching amount candidates. It is something.

[Effect]

In the pattern matching method according to the present invention, the matching amount candidate extraction circuit extracts some storage areas in the deviation memory in order from the storage areas having the highest data values as coordinate matching amount candidates, and stores these in the matching amount candidate memory. , among the combinations of these coordinate alignment candidates, the one with the maximum number of paired minutiae candidates after coordinate transformation is determined as the coordinate matching quantity.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings. 1st
In the figure, 1 to 5 are similar to the conventional one shown in FIG.
A pair of feature point candidate memory and a deviation memory, and 7 is a matching amount candidate extraction circuit connected to this deviation memory and extracting coordinate matching amount candidates in order from a storage area having a data value higher than that of the deviation memory 5; Reference numeral 8 denotes a matching gill candidate memory connected to the matching amount candidate extracting circuit 7 to store the extracted coordinate matching h (this candidate) and sending an output to the coordinate conversion circuit 2.

Next, the operation will be explained. Similar to the conventional pattern matching method, the feature points of the search pattern and the feature points of the file pattern are sequentially compared, and those whose pattern features have a difference within a predetermined threshold are selected as paired feature point candidates. The data in the storage area of the deviation memory 5 corresponding to the extracted positional differences ΔX, Δy and direction difference Δ0 is weighted. When the above-mentioned processing for all paired minutiae candidates extracted from the search pattern and file pattern is completed, the matching amount candidate extraction circuit 7 stores the address values in the X direction in order from the storage area with the highest data value in the deviation memory 5. Extract as a candidate for coordinate matching amount. The candidates for each coordinate matching amount at that time are assumed to be ΔXi (i=1.2°...n). In addition, candidates for the amount of coordinate matching in the y direction and further in the 0x direction are extracted in the same way, and each candidate for the amount of coordinate matching at that time is expressed as ΔY, (,j=1°,) 2 , - n LΔθ ( k=1,2....
no) k. These values are stored in the matching amount candidate memory 8.

These values are sent from the matching amount candidate memory 8 to the coordinate conversion circuit 2, and based on this, the coordinate conversion circuit 2 performs coordinate matching Il.
j as α=ΔXi, β=ΔY,, γ=Δθ.

, ] The position X+y and direction O of each feature point of the search pattern are subjected to coordinate transformation. The paired feature point candidate extraction circuit 3 extracts paired feature point candidates from this file pattern and the file pattern, and calculates the number of candidates. The number of paired feature candidates is the coordinate matching amount α=
Δxi (i=1+2+...n)+β
=ΔY-(,) = 1 + 2 +...nL
X, 1
yγ=Δθ (k=1
．． 2. . After the position Xt'/ and direction 0 of each feature point of the search pattern are coordinate transformed based on the coordinate matching amount determined by the coordinate transformation circuit 2, the search pattern and the file pattern are compared and matched. .

In the above embodiments, fingerprint pattern matching has been described, but the same effect can be achieved even if the invention is applied to matching other striped patterns.

〔Effect of the invention〕

As described above, according to the present invention, the combination in which the number of paired minutiae candidates after coordinate transformation is the maximum is determined as the coordinate matching amount from among the coordinate matching amount candidates. Even if the proportion of false pairs among them increases, the amount of coordinate matching between the search pattern and the file pattern can be determined accurately, which has the effect of improving the accuracy of pattern matching.

[Brief explanation of drawings]

FIG. 1 is a block diagram for explaining a pattern matching method according to an embodiment of the present invention, and FIG. 2 is a block diagram for explaining a conventional pattern matching method. 1 is a feature point memory, 2 is a coordinate conversion circuit, 3 is a pair feature point candidate extraction circuit, 4 is a pair feature point candidate memory, 5 is a deviation memory, 7 is a matching amount candidate extraction circuit, and 8 is a matching amount candidate memory. In addition, in the figures, the same reference numerals indicate the same or equivalent parts. Figure 1

Claims (1)

[Claims] In a pattern matching method in which a striped search pattern to be matched and a file pattern registered in advance are compared and matched using coded pattern features of pattern feature points, the pattern between the search pattern and the file pattern is A feature point whose features match within a predetermined threshold is set as a paired feature point candidate, and based on a positional difference between the paired feature point candidates, candidates for the amount of coordinate matching between the search pattern and the file pattern are extracted, respectively; Coordinate transformation is performed on each combination of coordinate matching amount candidates, and the combination of coordinate matching amount candidates that maximizes the number of paired minutiae candidates after the coordinate transformation is determined as the coordinate matching amount. Characteristic pattern matching method.