JPH0721822B2 - Fingerprint image center point detection method - Google Patents

Description

The present invention relates to a fingerprint image center point detection method that can be used as a reference for collation in a fingerprint collation method aimed at personal identification in a data terminal or the like.

The center point of the fingerprint image has been conventionally obtained by the following method. First, when the fingerprint image is binarized and then thinned, a thinned image formed by the core line 1 of the fingerprint ridge shown in FIG. 1 is obtained. With respect to these thinned images, (i) a point S ₀ above the thinned image is used as a start point, and (ii) one skeleton 1 descends from the point S _0, and the skeleton 1 is traced left and right to obtain an apex S. Ask for ₁ .

(Iii) Further down one core wire ₁ from the apex S ₁ , and obtain the apex S _{2 of the} core wire 1.

(Iv) Similarly, the vertices S ₂ , S ₃ ... Are obtained, and the vertices S _n when the core line 1 cannot be descended are used as the center points.

The above is the conventional method.

The method described above has the following two drawbacks. First, 2
It requires two pre-processings, value conversion and thinning. next,
This is a process that traces the thinning information locally, so the center point may not be found well due to divergence. There are two.

In order to eliminate these drawbacks, the present invention broadly calculates the normal direction of the fingerprint ridge from the original data of the fingerprint in the small area defined on the fingerprint image, and moves the small area along that direction. By doing so, the center point of the fingerprint image is determined. Hereinafter, the present invention will be described in detail.

The portion above the center point of the fingerprint image has a substantially concentric semicircular shape regardless of the type of fingerprint. Therefore, the center point of the fingerprint image is determined by moving the small areas defined in the upper part of the fingerprint image to the lower part one by one according to the normal direction of the fingerprint ridges in the small area. Can be found regardless of Therefore, it is effective to start tracking for detecting the center point from the upper part of the fingerprint image. less than,
First, the small area moving algorithm will be described, and then the center point determining algorithm will be described.

[Small area movement algorithm]

(A) Calculation of the normal direction of the fingerprint ridge and its reliability using Fourier transform The density distribution of a small area centered on the point (i ₀ , j ₀ ) in the coordinate system on the fingerprint image is f (x, y) (x, y = 0, 1, ..., L-1), the absolute value | F | of the Fourier transform F of the concentration distribution f obtained by the following equation (1) is used as the Fourier transform plane.

FIG. 2 shows an example of the density distribution f. In FIG. 2, 2 indicates a fingerprint ridge. In FIG. 2, the fingerprint ridge 2 is represented by a binary value, but in reality, a multivalued case is handled.

Further, FIG. 3 is a diagram in which | F | obtained from the concentration distribution f of FIG. 2 is modeled. In FIG. 3, 3 is a first peak point, 4 is a second peak point, z is a straight line connecting the first peak point 3 and the second peak point 4, θ is an angle formed by the straight line z with the u axis, and i is A point on the conversion surface, l is the first peak point 3 and the second peak point
The distance between the peak points 4, d _i is the distance between the straight line z and the point i,
r _i is the distance between the point i and the first peak point 3, and I ₂ is the second point
The value of the peak point 4, I _i , represents the value of the point i.

As shown in FIG. 3, the straight line z connecting the first peak point 3 located at the center of the conversion surface and having the maximum size and the second peak point 4 next to the first peak point 3 is the u-axis. Angle θ
Indicates the normal direction of the fingerprint ridge 2 in the small area.
The ambiguity A in this direction can be obtained by the equation (2).

Equation (2) is the point i with respect to the straight line z in FIG.
Is divided by the accumulation of moments when all the values of each point i are replaced by the value I ₂ of the second peak point 4, and represents the degree of blurring of the second peak point 4. The opacity A is normalized to 0A1 and α =
A value of 30 is 0.05 to 0.1 for an image with a clear direction and 0.2 to 0.4 for an unclear image. The exponential term in the equation (2) is a Gaussian probability density type weighting function, and the constant α is such that when the distance | l−r _i | until the value of this weighting function becomes 1 / e is Δl. , 1 has a physical meaning that the ratio of Δl to α is α (%).

(B) Detection of central coordinate point of new small area The central point of the new small area is determined from the straight line z and the opacity A obtained in (a). The A _t as a constant that represents the threshold of ambiguous degree A, in Figure 3, the center point of the new sub-region that case straight z of AA _t intersects the u = L / 2 or v = L / 2 To do. The coordinates of this point are (i ₁ , j ₁ ) in the coordinate system on the fingerprint image.

In the case of A> A _t The bisector of the line segment connecting the first peak point 3 and the point where the straight line z intersects u = L / 2 or v = L / 2 is set as the center point of the new small area. The coordinates of this point are (i ₁ ,
j ₁ ).

If A is greater than A _t are the case the normal direction of the ambiguous degree A of fingerprint ridge 2 obtained from the information of the Fourier transform plane is large, thus the purpose of performing a movement of more detailed subregion It was a different algorithm.

The above is the procedure for moving the small area from (i ₀ , j ₀ ) to (i ₁ , j ₁ ). The movement of the small area is continued in the same procedure below.

[Center point determination algorithm]

[Step 1] As shown in FIG. 4, an arbitrary point (i _a , j
Initialize _a ). The point (i _a , j _a ) is started, and the point moved by Δj (> 0) in the j direction until the value of the angle θ obtained by the above-mentioned small area moving algorithm becomes 0 is set as a new small area. Continue the process as the center of. The coordinates of the point where the value of the angle θ becomes 0 are (i ₀ , j ₀ ).

[Step 2] The small area is moved from (i ₀ , j ₀ ) according to the small area moving algorithm described above. At this time, the value of the opacity A corresponding to each small area center point is stored as A _n (n = 0, 1, ...). The movement ends when the i-coordinate of the small area center point decreases in the direction in which it decreases. The small area center point (i _n , j _n ) corresponding to _n where A _n is maximum is determined as the center point of the fingerprint image.

When the center point detection experiment of the fingerprint image is conducted by the above algorithm, the center point detection position error can be suppressed to about 1/4 of the small area for the fingerprint image of the same kind taken at another opportunity. It was confirmed. For example, for a fingerprint image with a resolution of about 50 microns, a size of 256 × 256, and each image having 256 gray levels, L = 32, or 32 × 32.
If having conducted by moving the small region, the error of each parameter indicated in the above algorithm as _{α = 30, A t = 0.15} ,
It is about 8 × 8 or 0.4 mm square.

In the above-mentioned algorithm, the movement of the small area when determining the center point is made horizontally from the upper left portion to the right by [Step 1], and downward rightward as [Step 2] when the angle θ reaches 0. Although it was going on, with a slight change, it is possible to move horizontally from the upper right to the left and then to the lower left. Both treatments have an equivalent effect.

Next, the construction and operation of an embodiment of the present invention will be described with reference to the block diagram of FIG.

In FIG. 5, a density distribution setting unit 11 receives fingerprint image information and calculates a density distribution f of a small area centered on the point (i ₀ , j ₀ ) in the coordinate system on the fingerprint image. Reference numeral 12 denotes a Fourier transform unit, which calculates the Fourier transform plane | F | from the density distribution f. 1
3 is a conversion surface processing unit, which converts the conversion surface | F |
The straight line z connecting the peak point and the opacity A are obtained. these
The above-mentioned "(a) Calculation of normal direction of fingerprint ridge and its reliability using Fourier transform" is executed in 11 to 13. Reference numeral 14 denotes a new small area setting unit, which determines the center point (i ₀ , j ₀ ) of the new small area from the straight line z obtained by the conversion surface processing unit 13 and the opacity A.
To decide. That is, the above-described "detection of the center coordinate point of a new small area" is executed. Thereafter, the center point (i _n , j _n ) of the small area is sequentially obtained.

Reference numeral 15 denotes a comparison unit, which compares the ambiguity A obtained each time by the conversion surface processing unit 13 with the opacity A obtained by the previous operation, and stores the larger one together with the coordinates of the center point at that time. Reference numeral 16 denotes a completion determining unit that determines that the movement is in a direction in which the coordinates of the small area center point decrease due to the movement, or when the movement is outside the fingerprint image, the end is determined, and the opacity A stored in the comparing unit 15 The center point coordinate of the largest one is output. At the same time, the input of fingerprint image information is stopped.

As described above, according to the present invention, the initial coordinate points are defined at the left and right lateral positions on the upper portion of the fingerprint image, and first, as the initial processing, the horizontal position is opposite to the left and right positions at which the initial coordinate points are defined. Calculate the normal direction of the ridge in the small area from the fingerprint ridge shape in the small area centered on the coordinate point, move the coordinate point horizontally until the normal direction becomes vertical,
Next, while moving the coordinate points in the calculated normal direction, similar to the initial processing, the normal direction of the ridge in the small area from the fingerprint ridge shape in the small area centered on the corresponding coordinate point, Furthermore, the ambiguity of the directionality that quantitatively represents the degree of reliability that can be specified in the normal direction is calculated, and the coordinate point movement in the normal direction is up to that point with respect to the vertical axis of the fingerprint image. When it becomes the opposite direction, or when the small area is outside the fingerprint image and the processing cannot be continued, the coordinate point with the smallest direction ambiguity until then is set as the center point of the fingerprint image. Therefore, there is an advantage that it does not require binarization and thinning processing, and that the flow in the direction of the fingerprint ridge is viewed in a global manner, so that it is unlikely to be affected by noise or the like. Further, the center point detecting ability of the method of the present invention is sufficiently practical, and can be applied to the setting of a coordinate system in fingerprint collation.

[Brief description of drawings]

FIG. 1 is an explanatory diagram of a conventional center point detection method, FIG. 2 is a diagram showing an example of a density distribution f of a small area on a fingerprint image, and FIG. 3 is a Fourier transform plane of the density distribution f of FIG. FIG. 4 is a diagram modeling F |, FIG. 4 is an explanatory diagram of the center point detection method of the present invention, and FIG.
The drawing is a block diagram showing the construction of an embodiment of the present invention. In the figure, 1 is the core line of the fingerprint ridge, 2 is the fingerprint ridge, 3 is the first peak point, 4 is the second peak point, 11 is the density distribution setting unit, 12 is the Fourier transform unit, and 13 is the conversion surface processing unit. , 14 is a new small area setting section, 15 is a comparing section, 16 is a completion judging section, S ₀ , S ₁ , S ₂ ...... S _n
Is the apex of the core line, z is a straight line connecting the first peak point and the second peak point, θ is the angle between the straight line z and the u axis, i is the point of the conversion surface, and 1 is the first peak point and the second peak point. , D _i is the distance between the straight line z and the point i, r _i is the distance between the point i and the first peak point,
I ₂ is the value of the second peak point, I _i is the value of the point i, (i _a , j _a ), (i
₀ , j ₀ ) and (i _n , j _n ) are the small area center coordinate points.

Claims (2)

[Claims] 1. An initial coordinate point is defined at a position on the upper left and right sides of the fingerprint image, and first, as an initial processing, the corresponding coordinate point is centered in the horizontal direction opposite to the left and right positions defining the initial coordinate point. Calculate the normal direction of the ridge in the small area from the fingerprint ridge shape in the small area, move the coordinate point horizontally until the normal direction is vertical, and then calculate the calculated normal direction. While moving the coordinate point to, the normal direction of the ridge in the small area from the fingerprint ridge shape in the small area centered on the corresponding coordinate point, as well as the normal direction, as in the initial processing. Calculate the degree of ambiguity that quantitatively expresses the degree to which it can be defined with the reliability of, and the movement of the coordinate points in the normal direction is opposite to that of the vertical axis of the fingerprint image, or When a small area is outside the fingerprint image and processing cannot continue , Most directional obscure degree less was coordinate point, the center point detecting method of a fingerprint image, characterized in that the center point of the fingerprint image before. 2. A normal direction of a fingerprint ridge existing in a small area,
And the ambiguity of the directionality are calculated using information of a conversion surface obtained by performing Fourier transform on the fingerprint ridge image in the small area. A method for detecting the center point of a fingerprint image.