TWI622001B - Finger vein identification method - Google Patents

Abstract

A finger vein identification method for solving the problem of poor recognition rate includes: calculating the difference between the sum of the gray scale values of the pixels of each of the adjacent X-axis image coordinates of the edge image in the Y-axis Determining whether the difference is not less than a threshold value. If the determination result is yes, one of the two adjacent X-axis image coordinates is used as a cutting point of the edge image, and the cutting point is along the edge image. The Y-axis extends to cut the edge image to generate a two-divided image; and the two-divided image respectively generates a histogram, and the histogram of the two-divided image is compared with the array control image to determine whether to compare the image with the array One of them is matched to produce a recognition result.

Description

Finger vein identification method

The present invention relates to a method for finger vein recognition, and more particularly to a method for identifying local features of a finger vein image.

Conventional authentication identity technology, when applied to technology products (such as notebook computers, financial cards or electronic wallets), mostly uses card and password identification mechanism to complete identification and authentication. However, for most people, the traditional authentication identity technology often has insufficient security and is prone to user problems. For example, when the card is lost, the password is forgotten, the password is recorded by an external program, etc., it will be correct. The user caused great inconvenience.

With the rapid development of science and technology, biometric technology (such as: face, retina, iris, fingerprint or finger vein) is replaced. Unlike the traditional authentication technology, it is based on the unique uniqueness of each individual. Features for identification and certification. The biometric technology can prevent unauthorized access through personal identification, while also preventing theft of personal mobile phones, ATM systems, personal computers, large workstations, computer network systems, and the like.

Among them, the conventional finger vein identification technology is based on the human body's tissue, which can display different optical characteristics under different wavelengths of light. When the finger skin is irradiated with near-infrared rays of uniform wavelength, the vein located in the lower layer of the human skin The blood vessel can be visualized and can be used as an identification feature. Then, a global feature conversion method is performed on the identification feature to achieve the purpose of identification. However, the global feature conversion method has the problem of declining the recognition rate. In view of this, the conventional finger vein identification technology still needs to be improved.

In order to solve the above problems, the present invention provides a finger vein identification method capable of identifying local features of a finger vein image.

The invention provides a finger vein identification method for authenticating an identification area, the method comprising: using an edge image as the identification area, and calculating a plurality of pixels of the edge image in the Y axis of different X-axis image coordinates The sum of the gray scale values is such that the X-axis image coordinates of the edge image respectively generate a gray scale quantity; calculate the difference of the gray scale quantities of any two adjacent X-axis image coordinates of the edge image, and determine whether the difference is not less than a threshold value, if the determination result is yes, one of the two adjacent X-axis image coordinates is used as a cutting point of the edge image, and the cutting point is cut along the Y-axis of the edge image to cut the edge The image E is used to generate a two-divided image. If the result of the determination is negative, and the difference between the gray scales of any two adjacent X-axis image coordinates of the edge image is less than the threshold value, the threshold value is lowered and re-determined. Whether the difference is not less than the threshold value; and generating a histogram for each of the two divided images, the histogram of the two divided images is respectively compared with the array control image, and determining whether the image is compared with the array One consistent, to produce a recognition result.

Accordingly, the finger vein identification method of the present invention can identify the local features of the finger vein image by using the difference in brightness between the bone and the muscle at the joint of the user's finger as a condition for segmentation. Thereby, the present invention can achieve the effect of improving the overall recognition rate.

The light source is illuminated toward a user's finger and captured by an image capture device to generate an original image, and an edge detection is performed on the original image to generate the edge image. In this way, the time required for image processing can be reduced, and the overall recognition efficiency can be improved.

Wherein, a grayscale processing is performed on the original image. In this way, the time required for image processing can be reduced, and the overall recognition efficiency can be improved.

The noise is removed by a median filter on the original image. So, make this The original image is smoother and has the effect of improving the accuracy of subsequent image processing.

Among them, a light source of near-infrared light having a wavelength of between 700 nm and 1000 nm is irradiated toward the finger of the user. In this way, the original image captured can be made clearer and has the effect of improving the original image resolution.

The edge image is sequentially subjected to morphological image processing of erosion and expansion. In this way, the effect of eliminating noise can be achieved, and the efficiency of subsequent image processing can be improved.

Wherein, a center line is set for the edge image, and the center line divides the edge image into an upper half and a lower half, and searches for the maximum of the Y-axis image coordinates of the plurality of white pixels in the upper half of the edge image. And searching for the smallest of the Y-axis image coordinates of the plurality of white pixels in the lower half of the edge image, the largest and the smallest of the Y-axis image coordinates extending along the X-axis to the boundary of the edge image, A subsequent image processing procedure is performed to generate an area of interest as the identification area. In this way, the time required for image processing can be reduced, and the overall recognition efficiency can be improved.

The region of interest is generated by a normalization process by a nearest neighbor method. In this way, the time required for image processing can be reduced, and the overall recognition efficiency can be improved.

The histogram of the two-divided image is compared with the array control image by an Euclidean distance method. In this way, it has the effect of improving the overall recognition rate.

The edge image is subjected to brightness compensation by normalization of a gray scale. In this way, it has the effect of improving the image recognition rate.

〔this invention〕

S1‧‧‧ capture procedure

S2‧‧‧Detection procedure

S3‧‧‧ cutting procedure

S4‧‧‧ comparison procedure

S5‧‧‧Optimizer

S51‧‧‧ Gray Steps

S52‧‧‧ Smoothing steps

S6‧‧‧ calibration procedure

S61‧‧‧ Morphological steps

S62‧‧‧Feature Search Step

S63‧‧‧ formalization steps

S64‧‧‧Light compensation step

C‧‧‧Image capture device

CL‧‧‧ midline

E‧‧‧Edge image

F‧‧‧ finger

L‧‧‧Light source

O‧‧‧ original image

P‧‧‧ split point

R‧‧‧ Identification area

S‧‧‧ split image

Figure 1 is a flow chart showing the method of an embodiment of the present invention.

Fig. 2 is a schematic view showing the finger vein of an embodiment of the present invention.

Figure 3a is a schematic diagram of an original image of an embodiment of the present invention.

Figure 3b is a schematic diagram of image segmentation according to an embodiment of the present invention.

Figure 3c is a schematic view of a region of interest in accordance with an embodiment of the present invention.

Figure 4 is a schematic diagram of an optimization procedure of an embodiment of the present invention.

Fig. 5 is a view showing a calibration procedure of an embodiment of the present invention.

The above and other objects, features and advantages of the present invention will become more <RTIgt;

The "Color Level" as used throughout the present invention refers to the degree of gradation of the color component or brightness exhibited by the pixel, for example, the red (R), green (G), and blue (B) components of the color image. The range of the gradation ranges from 0 to 255; alternatively, the range of the gradation of the grayscale image (Luminance) may range from 0 to 255, which can be understood by those of ordinary skill in the art to which the present invention pertains.

Please refer to FIG. 1 , which is a schematic flowchart of a method for detecting a vein identification method according to the present invention. The method includes: a sampling program S1, a detecting program S2, a cutting program S3, and a matching program S4. The method is used to authenticate an identification area R, which includes a vein image of a finger F of a user.

Referring to FIG. 2 together, the capturing program S1 can illuminate the user's finger F with a light source L, and then take an image capturing device C (eg, a camera) to generate an original image. Can be as shown in Figure 3a. The light source L may be infrared light, and is preferably near-infrared light having a wavelength of between 700 nanometers (nm) and 1000 nanometers. In this way, the original image O can be made clearer and has the effect of improving the accuracy of subsequent image processing.

The detection program S2 can detect the edge feature of the original image O by an edge search method to generate an edge image E, and use the edge image E as the identification area. R performs a subsequent image processing procedure, and the image size of the recognition area R is not larger than the image size of the edge image E, as shown in FIG. 3b. For example, the recognition area R can be preset to be equal to the image size of the edge image E. The edge image E is a two-dimensional image, and the edge search method may be a gradient operation sub-edge search method, such as: Canny Edge Detection or Sobel Edge Detection. In this embodiment, the Sobel edge detection is used, which can be understood by those having ordinary knowledge in the technical field of the present invention, and will not be further described herein.

The cutting program S3 is capable of cutting the edge image E to generate at least two divided images S. In detail, the cutting program S3 calculates the difference between the sum of the gray scale values of the pixels of the two adjacent X-axis image coordinates of the edge image E in the Y-axis. Then, it is determined whether the difference is not less than a threshold value. If the determination result is yes, one of the two adjacent X-axis image coordinates is used as a cutting point P, preferably the one with a larger gray scale quantity is used as the Cutting the edge P of the edge image E, and cutting the edge image E along the Y-axis of the edge image E to generate the two-divided image S, as shown in FIG. 3b; if the determination result is no And if the difference between the gray scales of any two adjacent X-axis image coordinates of the edge image E is less than the threshold value, the threshold value is lowered, and it is re-determined whether the difference is not less than the threshold value. In the present embodiment, the threshold value is used to distinguish the function between the bone and the muscle at the joint of the finger F. Therefore, when the threshold value can reach the function, the threshold value is not subject to any restriction.

The grayscale value may also be generated by calculating an average of the grayscale values of the plurality of pixels of the edge image E in the Y-axis of different X-axis image coordinates. In this way, the interference of the noise can be reduced, and the invention has the effect of improving the accuracy of the subsequent image processing.

The comparison program S4 can count the grayscale values of the divided images S to respectively generate a histogram image. Subsequently, each of the histogram images is compared with an array of image images stored in a database by an identification technique (each group of control images is a histogram image of a segmented image of a user's finger vein image), To judge whether the identification is successful Or failed. Each of the control images is a histogram image that is established in advance through the capture program S1, the detection program S2, and the cutting program S3. The identification technique may be an Euclidean Distance method. It will be understood by those of ordinary skill in the art to which the present invention pertains, and no further description is provided herein.

Referring to FIG. 1 and FIG. 4 together, the vein identification method of the present invention can further include an optimization program S5, which can be used to enhance the recognition rate of the original image O, and can include a gray scale. Step S51 and a smoothing step S52, wherein if the original image O is an RGB image, the grayscale step S51 can perform grayscale processing on the original image O to calculate each image pixel of the original image O. Brightness. In detail, the gray-scale step S51 converts the hue of the original image O to the brightness of the color-scale range from 0 to 255 according to the gradation of the red, green, and blue components of each pixel of the original image O. In this way, the time required for image processing can be reduced, and the overall recognition efficiency can be improved. In addition, when the color gradation of the original image O is in the range of 0 to 255 (that is, the original image is a grayscale image), the grayscale step S51 may be omitted, which is in the technical field of the present invention. Those with ordinary knowledge can understand.

The smoothing step S52 can remove the noise of the original image O by using an image filter, so that the original image O is more advantageous for subsequent image processing analysis. For example, the image filtering may be a low pass filter, a median filter, or a high pass filter. In this embodiment, the image filtering system is medium. The value filter can also make the original image O smoother while eliminating the pepper salt noise of the original image O. Thus, the present invention has the effect of improving the accuracy of subsequent image processing. In addition, when the resolution of the original image O does not affect the subsequent image processing analysis, the smoothing step S52 may be omitted.

Referring to Figures 1 and 5, wherein the vein identification method of the present invention can further comprise a calibration program S6, the calibration program S6 can include a morphological step S61, a special Steps of searching for step S62, a normalizing step S63, and a light compensation step S64. The morphological step S61 performs Morphology processing on the edge image E. In detail, the more common operations in morphology can be divided into: Dilation, Erosion, Opening, Closing, and Region Filling, etc., in this embodiment, The morphological step S61 performs the operations of erosion and expansion on the edge image E in order to achieve the effect of eliminating noise. Thus, the present invention has the effect of improving the accuracy of subsequent image processing. Moreover, when the resolution of the edge image E does not affect the subsequent image processing analysis, the morphological step S61 can be omitted.

The feature searching step S62 selects a Region of Interest (ROI) as the recognition region R for the edge image E to perform a subsequent image processing procedure. In detail, the region of interest is a rectangular image in the image of the edge of the finger in the edge image E, as shown in Figure 3c. The region of interest is generated by dividing the edge image E into an upper half and a lower half by using the edge image E and the center line CL as a boundary, and searching for a plurality of white pixels of the edge image E in the upper half ( The largest of the Y-axis image coordinates with a grayscale value of 255), and the smallest of the Y-axis image coordinates of the plurality of white pixels searching for the edge image E in the lower half, the largest and smallest of the Y-axis image coordinates Each extends along the X-axis to the boundary of the edge image E to produce the region of interest.

In addition, the size of the region of interest may be slightly different each time the detection procedure S2 is executed, and therefore, the size of the region of interest can be made uniform by the normalization step S63. In detail, the normalization step S63 can normalize the edge image E by a nearest neighbor method (Nearest Neighbor, NN). For example, an edge image E of 180 x 60 pixels is equalized by a ratio of a human finger aspect ratio of 3:1. In this way, the time required for image processing can be reduced, and the overall recognition efficiency can be improved.

In addition, since the finger F of the user is photographed, the brightness of the external environment is affected, and therefore, the edge images E generated after each shooting have a difference in brightness and darkness. Therefore, the light compensation step S64 can perform brightness compensation on the edge image E by a grayscale normalization to reduce the problem of uneven light, and has the effect of improving the image recognition rate. For example, when the brightness of the edge image E is too bright, the sum of the gray scale values of the pixels of the X-axis image coordinate of the edge image E in the Y-axis is the smallest, as the Y of the gray-scale distribution map. The starting point of the axis.

In summary, the finger vein identification method of the present invention can identify the local features of the finger vein image by using the difference in brightness between the bone and the muscle at the joint of the user's finger as a segmentation condition. Thereby, the present invention can achieve the effect of improving the overall recognition rate.

While the invention has been described in connection with the preferred embodiments described above, it is not intended to limit the scope of the invention. The technical scope of the invention is protected, and therefore the scope of the invention is defined by the scope of the appended claims.

Claims (10)

A finger vein identification method for authenticating an identification area, the method comprising: using an edge image as the identification area; calculating any two adjacent X-axis image coordinates of the edge image, each of which is in the Y axis a difference between the sum of the grayscale values of the pixels; determining whether the difference is not less than a threshold value, and if the determination result is yes, using one of the two adjacent X-axis image coordinates as a cutting point of the edge image, and The edge image E is cut along the Y-axis of the edge image to generate a two-divided image. If the determination result is no, and the grayscale amount of any two adjacent X-axis image coordinates of the edge image is different, If the threshold value is less than the threshold value, the threshold value is lowered, and whether the difference is not less than the threshold value is re-determined; and the two-part image is generated by a histogram, and the histograms of the two-divided images are respectively performed with the array control image. The comparison determines whether it matches one of the array control images to generate a recognition result. The finger vein identification method according to claim 1, wherein a light source is irradiated toward a user's finger and photographed by an image capturing device to generate an original image, and an edge is performed on the original image. Detect to generate the edge image. The finger vein identification method according to claim 2, wherein the light source of near-infrared light having a wavelength of between 700 nm and 1000 nm is irradiated toward the finger of the user. The finger vein identification method of claim 2, wherein the grayscale processing is performed on the original image. The finger vein identification method of claim 4, wherein the original image is subjected to a median filter to remove noise. The finger vein identification method according to claim 1, wherein the edge image is sequentially subjected to morphological image processing of erosion and expansion. The finger vein identification method of claim 1, wherein a center line is set for the edge image, and the center line separates the edge image into an upper half and a lower half, and the edge image is searched for The largest of the Y-axis image coordinates of the plurality of white pixels in the upper half, and the smallest one of the Y-axis image coordinates of the plurality of white pixels searching for the edge image in the lower half, the largest of the Y-axis image coordinates and The smallest ones each extend along the X-axis to the boundary of the edge image to generate a region of interest as the recognition region to perform subsequent image processing procedures. The finger vein identification method according to claim 7, wherein the region of interest is generated by a normalization process by a nearest neighbor method. The finger vein identification method according to claim 1, wherein the histogram of the two-divided image is compared with the array control image by an Euclidean distance method. The finger vein identification method of claim 1, wherein the edge image is subjected to brightness compensation by normalization of a gray scale.