CN110119724A - A kind of finger vein identification method - Google Patents

Abstract

The invention relates to a finger vein identification method, comprising the following steps: S1, performing ROI positioning on an image collected by a high-definition camera; S2, performing grayscale normalization processing on different parts of the image; S3, performing mean filtering and Gauss filtering on the image Eliminate noise; S4, use curvature detection to determine the center position of the texture; S5, enhance the contrast of the grayscale image through gamma transformation; S6, use the refinement algorithm based on the index table to determine the edge pixels; S7, extract the finger vein skeleton for fusion , Dilation, Smoothing. Finger vein recognition is performed by template matching, Hu invariant moments, and improved Zernike moments. Especially under the condition of unclear image acquisition, the method has strong anti-interference ability and high-precision recognition ability.

Description

A method of finger vein recognition

technical field

The invention belongs to the technical field of image processing and biometric identification, and in particular relates to a finger vein. recognition methods.

Background technique

Finger vein recognition is a technology that uses human biometrics to identify individuals. At this stage, biometric identification technology mainly relies on fingerprint identification and iris identification, and these two types of body surface information are collected from the human body, which are easy to be stolen and imitated. Finger vein recognition technology, as a new generation of human body feature recognition technology, prevents personal information features from being stolen due to its liveness and uniqueness. At the same time, it also avoids the impact of other environmental factors.

During the identification process of finger vein identification equipment, the original finger vein images collected are of low definition due to the inability to accurately locate the core components (light source, filter, camera); the performance of the identification algorithm is easily affected by image rotation, translation and noise. influences.

Therefore, a finger vein recognition algorithm that preprocesses the original image and effectively suppresses the influence of image rotation and translation on the recognition accuracy needs to be developed urgently.

SUMMARY OF THE INVENTION

The present invention provides a finger vein identification method.

Finger vein identification method mainly includes four parts, image preprocessing, finger vein feature extraction to generate matching template, and finger vein algorithm identification.

The image preprocessing part consists of image grayscale, ROI region extraction, linear normalization, mean filter, Gaussian filter and so on. The ROI area is extracted by the Sobel operator, and the grayscale image is denoised and smoothed by linear filtering.

Feature extraction is to extract clear finger vein patterns and skeletons from the grayscale image after image preprocessing, and finally generate a registration template. The invention adopts the method of curvature detection to detect the extreme point of grayscale change in the finger vein image, and extracts the finger vein pattern from the more complex background; then the image quality is enhanced by the change of Gamma, and the OTSU threshold segmentation method is used to separate the The grayscale image is binarized, and through the method of neighborhood detection, the thicker veins after binarization are refined, and the finger vein skeleton is extracted. Finally, the finger vein skeleton is processed by rotation fusion, expansion and other methods to generate a finger vein template for matching.

The identification part of finger vein algorithm includes template matching, Hu invariant moment and improved Zernike moment matching. Template matching is an important part of the matching algorithm because of its fast processing speed. When the image acquisition quality is high, the recognition can be performed directly to improve the recognition efficiency. When the template matching cannot accurately identify, the seventh-order Hu moments of the collected finger vein images and the samples in the finger vein library are compared and calculated for identification. Further, by calculating the improved Zernike seventh-order moment of the finger vein image, it can make up for the difficulty of Hu's recognition in high order due to rotation.

When the matching degree of any one of the three recognition functions of template matching, Hu moment and Zernike moment is higher than the threshold, it is determined that the matching is successful.

Compared with the prior art, the beneficial effects of the present invention are as follows:

The invention can realize the high-precision acquisition of the finger vein, and solve the identification difficulty caused by image rotation and translation in the identification process. 150 finger vein samples were collected from 50 individuals, under the condition that the finger rotates at a maximum left-right angle of 30°. The identifications are listed in the table below.

The advantage of the present invention lies in that, in the process of finger vein identification, due to the different postures and lighting of the fingers. It is difficult to ensure that the template collected by the same individual during finger vein entry is consistent with the valid information extracted during matching. Most of them will be rotated and offset. This brings a lot of challenges to high-precision recognition. Based on this, the present invention proposes an improved Zernike moment recognition algorithm, so as to better characterize the global features of finger vein patterns and improve the recognition accuracy.

Description of drawings

Fig. 1 is the flow chart of finger vein image feature extraction according to the present invention

Fig. 2 is the flow chart of the finger vein recognition algorithm of the present invention

Detailed ways

The invention provides a method for identifying human finger veins with accurate identification.

The specific embodiments of the present invention are described below so that those skilled in the art can understand the present invention.

Referring to FIG. 1 , FIG. 1 is a flowchart of feature extraction of an embodiment of a finger vein recognition algorithm; as shown in FIG. 1 , the extraction method includes steps S1 to S7 .

In step S1, the currently collected finger vein image is obtained, the edge in the vertical direction is detected by calculating the gradient in the horizontal direction of the grayscale image, and the edge in the horizontal direction is detected by calculating the gradient in the vertical direction of the grayscale image;

In step S2, the gray value of its effective information is more concentrated, and the effective interval span is less than 0-255, using the formula Perform linear normalization;

In step S3, the formula Perform mean filtering, Gaussian filtering using the formula

In step S4, the pixel gray value of the processed image is derived to determine the center position of the finger vein pattern;

In step S5, use the formula s=cr ^γ to perform a Gamma transform operation on the gray value of the image, so as to distinguish and enhance the gray level of the image;

In step S6, by traversing the edge of the finger vein binary image, and according to the index table to determine the situation of the eight fields of the point. If the result is 1, it is removed; if the result is 0, it is an image edge pixel and needs to be retained.

In step S7, a pair of finger vein skeletons are rotated by multiple angles, and then multiple skeleton images are fused. The fused finger vein image is then subjected to morphological expansion processing using the structural elements whose computing kernel is a square. Generate a finger vein registration template. where the computing core is

Referring to FIG. 2 , FIG. 2 shows the identification process of the finger vein algorithm.

Further, the template matching formula is T=1-∫∫ _s (ft) ² dx dy, where S represents the definition domain of the image t(x, y), and the similarity S, 1 of the finger vein input image and the sample image is calculated Indicates the highest similarity, and 0 means no similarity.

Further, the calculation method of Hu moment is as follows

φ ₁ =η ₂₀ +η ₀₂

φ ₂ =(η ₂₀ -η ₀₂ ) ² +4η ₁₁ ²

φ ₃ =(η ₃₀ -3η ₁₂ ) ² +(3η ₂₁ -η ₀₃ ) ²

φ ₄ =(η ₃₀ +η ₁₂ ) ² +(η ₂₁ +η ₀₃ ) ²

φ ₅ =(η ₃₀ -3η ₁₂ )(η ₃₀ +η ₁₂ )[(η ₃₀ +η ₁₂ ) ² -3(η ₂₁ +η ₀₃ ) ² ]+

(3n ₂₁ -n ₀₃ )(n ₂₁ +n ₀₃ )[3(n ₃₀ +n ₁₂ ) ² -(n ₂₁ +n ₀₃ ) ² ]

φ ₆ =(η ₂₀ -η ₀₂ )[(η ₃₀ +η ₁₂ ) ² -(η ₂₁ +η ₀₃ ) ² ]+

4n ₁₁ (n ₃₀ +n ₁₂ ) (n ₂₁ +n ₀₃ )

φ ₇ =(3η ₂₁ -η ₀₃ )(η ₃₀ +η ₁₂ )[(η ₃₀ +η ₁₂ ) ² -3(η ₂₁ +η ₀₃ ) ² ]+

(3n ₁₂ -n ₃₀ )(n ₂₁ +n ₀₃ )[3(n ₃₀ +n ₁₂ ) ² -(n ₂₁ +n ₀₃ ) ² ]

are the seven invariant moments constructed for the second and third order moments of Hu, and η _pq is the (p+q) order invariant moment.

Further, the calculation method of the improved Zernike moment is as follows.

One is to find the shape centroid O(x _o , y _o ) of the finger vein vein, and then use the Euclidean distance to find the pixel point B(x _b , y _b ) on the vein that is farthest from the shape centroid O, and r is both The distance between them, thus determining the circumscribed circle of its finger veins, the radius is r, and all the target points are normalized into the unit circle, which makes the Zernike moment obtained have translation and scale invariance;

The second is to calculate the 0th order geometric moment of the finger vein venation in the image.

m ₀₀ =∫∫f(x,y)dxdy

The third is to calculate the Zernike moments of each order in the unit circle

The fourth is to normalize the Zernike moment using the 0th order geometric moment

Further, the eigenvalues of the finger vein image are compared with the finger vein database, when the matching degree of template matching, seventh-order Hu moment value, and improved Zernike moment value is compared with the threshold, any one of If the value is greater than the threshold, the match is successful.

Claims (5)

1. a kind of finger vein identification method, which comprises the steps of:

S1, ROI positioning is carried out to the image of high-definition camera acquisition；

S2, gray scale normalization processing is carried out to image different piece；

S3, mean filter, gaussian filtering elimination noise are carried out to image；

S4, lines center is judged using curvature measuring；

S5, the contrast that enhancing grayscale image is converted by gamma；

S6, using judging edge pixel point based on the thinning algorithm of concordance list；

S7, extraction is referred to vein skeleton merged, expanded, smoothing processing.Pass through template matching, Hu not bending moment, improved Zernike square carries out finger hand vein recognition.

2. finger vein identification method according to claim 1, which is characterized in that use Sobel operator pair in step s 2 Gray level image carries out edge detection.

3. finger vein identification method according to claim 1, which is characterized in that finger vein pattern extraction process mean curvature Detection, power law transformation, the thinning algorithm based on concordance list.

4. finger vein identification method according to claim 1, which is characterized in that improved when carrying out finger hand vein recognition Zernike square.Its calculation method is as follows.

Find out the shape mass center O (x for referring to vein train of thought_o, y_o), recycle Euclidean distance to find out in its train of thought apart from shape mass center O most Remote pixel B (x_b, y_b), r is distance between the two, so that it is determined that it refers to the circumscribed circle of vein train of thought, radius r, institute Some target points normalize in unit circle, this Zernike square allowed for has translation and scale invariance；

Calculate 0 rank geometric moment of image middle finger vein train of thought.

m₀₀=∫ ∫ f (x, y) dxdy

Each rank Zernike square in unit of account circle

Zernike square is normalized using 0 rank geometric moment

5. finger vein identification method according to claim 1, it is characterised in that: during referring to hand vein recognition, by In posture difference, illumination difference that finger is put into.It is difficult to ensure that same individual is referring to the template and matching acquired when vein typing When the effective information that extracts be consistent.It is most of all to rotate, deviate.This brings no small choose to high-precision identification War.Based on this, the invention proposes a kind of improved Zernike square recognizers, carry out better characterize with this and refer to veinprint Global characteristics improve accuracy of identification.