CN110163119A - A kind of finger vein identification method and system - Google Patents

Abstract

The invention discloses a finger vein recognition method and system, comprising: determining the upper and lower edge point sets of the finger area, and thinning the edge to a pixel width; Edge point set; according to the obtained pixel point coordinates, correct the rotation of the finger, and set the background gray value to 0; obtain the ROI, select the width as 0.73 times the original image, select the bottom edge coordinates when intercepting the upper edge, and select the lower edge When intercepting, select the uppermost edge coordinates for height interception; perform histogram equalization and Gabor filtering on the ROI; extract ROIs from a total of 3816 636 types of vein images and save them; use the one vs n method for matching, and use zero-mean normalization Calculate the similarity score of two ROI images in the direction of normalization, ranging from 0 to 1. The closer the value is to 1, the higher the similarity. The subject of the image with the highest similarity to the image to be matched is judged as the subject of the image to be matched. .

Description

A finger vein recognition method and system

technical field

The invention relates to the technical fields of image processing and biological feature recognition, in particular to a finger vein recognition method and system.

Background technique

Finger vein recognition is a kind of biometric recognition technology. Finger vein recognition technology is based on the characteristics of blood flow that can absorb characteristic wavelength relationships. Using near-infrared light to irradiate fingers, images of invading finger veins can be captured. Because the characteristics of finger veins are difficult to be copied, and the characteristics of each person's finger veins are different, and they will hardly change with age. Therefore, finger vein recognition technology has the characteristics of living body recognition, high security, and uniqueness. , Hotel management, government agencies, prison access control, medical appraisal, etc. have great application prospects.

In the process of finger vein recognition or verification, due to unstable illumination when collecting vein images, the rotation of fingers may lead to uneven image quality. Therefore, an algorithm that is more robust to errors caused by illumination and finger rotation is needed. Finger vein recognition can be applied in real life.

Contents of the invention

The technical problem to be solved by the present invention is to provide a finger vein recognition method and system to solve the above problems in view of the shortcomings of current finger vein recognition algorithms.

A finger vein recognition method, comprising:

S1. Determine the upper and lower edge point sets of the finger area in the original image, refine the edge to a pixel width, and place the finger area in the original image horizontally;

S2. Select suitable points from the thinned edge point set to perform edge extension to obtain a real edge point set;

S3. Correct the finger rotation, and set the gray value of the non-finger area to 0 according to the real edge point set;

S4, the image processed through S3 is cropped, and the width selection is 0.73 ± 5% times of the original image, preferably 0.73 times, and the bottom edge coordinates are selected when the upper edge is intercepted, and the uppermost edge coordinates are selected when the lower edge is intercepted. Highly intercepted to obtain the ROI of the vein region of interest;

S5. Perform histogram equalization and Gabor filtering on the ROI of the vein ROI to obtain the ROI of the vein ROI after image enhancement, to be used for matching;

S6. Using the steps of S1-S5 to process and extract vein ROIs for multiple preset categories of vein images and save them;

S7. The one vs n method is adopted for vein identification and matching, and the similarity score of each two vein ROI images is calculated using the zero-mean normalization method, and the vein ROI belonging to the same category is identified according to the similarity scores image.

Further, in step S1, specifically include:

S11. Obtain the edge point set a, if the difference between the gray value of the current pixel point and the gray value of the pixel point 2 coordinates above it is more than 33, then the pixel point is determined to be an edge point;

S12. Obtain the edge point set b, and use the Sobel operator to calculate the gradient of the entire image. When the current pixel gradient exceeds the gradient value of two adjacent pixels in the gradient direction, the pixel is determined to be an edge point;

S13. Perform an intersection operation on the edge point set a and the edge point set b to obtain the required edge point set, but only keep the first 15 pixels in the frequency of the abscissa, and refine the edge, that is, there are at most an upper edge point and a lower edge point;

S14. The edge point set is represented in the image as a horizontal curve with uneven thickness, and thinning is performed on it, and the edge is thinned to a width of one pixel.

Further, in step S2, in the set of upper and lower edge points, respectively select a point whose ordinate is closest to the center as the starting point to expand the edge, set the point (x, y), if the three adjacent coordinates are extended to the left of the point (x-1, y-1), (x-1, y), (x-1, y+1) has a gray value of 255, then set this adjacent point as an edge point and continue to expand to the left , if none of the grayscale values is 255, then take the point with the largest gradient calculated in S1 of these three coordinates as the edge point, and obtain the complete finger outline through continuous expansion.

Further, in step S3, select four edge points (x1, y1), (x2, y1), (x3, y2), (x4, y2) at the width of 0.23 and 0.77, and calculate the angle of finger rotation: Rotate the image to the horizontal according to the calculated rotation angle, and set the background gray value to 0.

Further, in step S4, specifically include:

S41. Use a rectangular window with a width of 50, move to the right from the middle coordinate of the vein, calculate the average gray value of the window every time you move a coordinate, and return the 5 window coordinates with the largest average gray value, and select the coordinates from these 5 The smallest one is used as the baseline of the ordinate, and the vein region of interest ROI of 0.73 times the width of the original image is intercepted to the left;

S42. After the width interception is completed, perform height interception, select the bottom edge coordinates for the upper edge point set of the image, and select the uppermost edge coordinates for the lower edge point set to perform height interception to obtain the ROI of the vein ROI.

Further, in step S5, specifically include:

S51. Perform histogram equalization on the intercepted ROI, specifically, use a limited contrast adaptive histogram equalization algorithm;

S52. Perform Gabor four-direction filtering of 0°, 45°, 90°, and 135° on the ROI after the histogram equalization.

Further, in step S7, specifically include:

S71. Firstly, classify all vein images;

S72. Select one as the image to be matched, divide the image to be matched and the candidate image into 8 rectangular areas, calculate the ZNCC value of the rectangular area corresponding to each pair of positions, and calculate the average ZNCC value of the two images as the similarity measurement value, ZNCC The calculation requires the sum of the square differences of the two images, their respective grayscale averages, and standard deviations. Let the size of the two corresponding rectangular areas be (2n+1)×(2n+1), and the center coordinates are (u ₁ ,v ₁ ),(u ₂ ,v ₂ ), the sum of square differences is The average gray value is: The standard deviation is: ZNCC is: The average ZNCC value of 8 rectangular areas is taken as the similarity measure standard. The ZNCC value ranges from 0 to 1. The closer the value is to 1, the higher the similarity is. The class of the image with the highest similarity to the image to be matched is judged as The category of the image to be matched;

S73. Calculate and sort the similarity between the image to be matched and the remaining images in the database, select the class of the image with the largest similarity as the predicted class, and compare it with the actual class. If they are equal, the prediction is correct, otherwise the prediction is wrong;

S74: The number of correct predictions during matching/total number of matches is the accuracy rate of the algorithm.

A finger vein recognition system, comprising: a processor and a storage device; the processor loads and executes instructions and data in the storage device to implement any one of the finger vein recognition methods described in claims 1-7.

The beneficial effect of the present invention is that: the finger vein recognition method adopts the method of edge expansion, can extract the ROI of the vein pattern very accurately, and at the same time corrects the possible rotation of the finger during the image acquisition process, and adopts a histogram The equalization method makes the image more robust to the light during the matching process. The four-directional Gabor filter is used to enhance the texture characteristics of the veins. Using the ZNCC image matching method, the finger vein database released by Shandong University is used. An accuracy rate of 98.7% was achieved.

Description of drawings

The present invention will be further described below in conjunction with accompanying drawing and embodiment, in the accompanying drawing:

Fig. 1 is the ROI extraction flowchart of finger vein recognition method of the present invention;

Fig. 2 is the authentication matching flowchart of finger vein recognition method of the present invention;

Fig. 3 is the changing process of the vein image when the ROI of the finger vein recognition method of the present invention is extracted;

Fig. 4 is the equal error rate curve of the finger vein recognition method of the present invention;

Fig. 5 is the accuracy rate curve of the finger vein recognition method of the present invention.

Detailed ways

In order to make the purpose, technical solution and advantages of the present invention clearer, the embodiments of the present invention will be further described below in conjunction with the accompanying drawings.

As shown in Figure 1 and Figure 2, a finger vein recognition method comprises the following steps:

S1. Determine the upper and lower edge point sets of the finger area in the original image, refine the edge to a pixel width, and place the finger area in the original image horizontally;

The specific process is as follows:

S11. Obtain the edge point set a. If the difference between the gray value of the current pixel point and the gray value of the two pixels above it exceeds 33, the pixel point is determined to be an edge point.

S12. Obtain the edge point set b, and use the Sobel operator to calculate the gradient of the entire image. When the gradient of the current pixel exceeds the gradient values of two adjacent pixels in the gradient direction, the pixel is determined as an edge point.

S13. Perform an intersection operation on the edge point set a and the edge point set b to obtain the required edge point set, but only keep the first 15 pixels in the frequency of the abscissa, and refine the edge, that is, there are at most One upper edge point and one lower edge point.

S14. The edge point set is represented in the image as a horizontal curve with uneven thickness, and thinning is performed on it, and the edge is thinned to a width of one pixel.

S2. Select suitable points from the thinned edge point set to perform edge extension to obtain a real edge point set;

The specific process is as follows:

In the set of upper and lower edge points, select a point whose ordinate is closest to the center as the starting point to expand the edge. Let this point (x, y) expand to the left of the point if the adjacent three coordinates (x-1, y- 1), (x-1, y), (x-1, y+1) has a gray value of 255, then set this adjacent point as an edge point, and continue to expand to the left, if there is no gray value is 255, the point with the largest gradient calculated in S1 by these three coordinates is taken as the edge point, and the complete finger outline is obtained by continuous expansion.

S3. In order to reduce the possibility of the finger rotation leading to a drop in accuracy during image acquisition, the finger rotation is corrected, and the gray value of the non-finger area is set to 0 according to the real edge point set;

The specific process is as follows:

Select four edge points (x1, y1), (x2, y1), (x3, y2), (x4, y2) at 0.23 width and 0.77 width, and calculate the angle of finger rotation: Rotates the image horizontally according to the calculated rotation angle. And set the background gray value to 0.

S4, the image processed through S3 is cropped, and the width selection is 0.73 ± 5% times of the original image, preferably 0.73 times, and the bottom edge coordinates are selected when the upper edge is intercepted, and the uppermost edge coordinates are selected when the lower edge is intercepted. Highly intercepted to obtain the ROI of the vein region of interest;

The specific process is as follows:

S41. Use a rectangular window with a width of 50, move to the right from the middle coordinate of the vein, calculate the average gray value of the window every time you move a coordinate, and return the 5 window coordinates with the largest average gray value, and select the coordinates from these 5 The smallest one is used as the baseline of the ordinate, and the ROI of 0.73 times the width of the original image is intercepted to the left.

S42. Perform height interception after the width interception is completed, select the bottom edge coordinates for the upper edge point set of the image, and select the uppermost edge coordinates for the lower edge point set to perform height interception to obtain a vein region of interest (ROI).

S5. Perform histogram equalization and Gabor filtering on the ROI of the vein ROI to obtain the ROI of the vein ROI after image enhancement, to be used for matching;

The specific process is as follows:

S51. Perform histogram equalization on the intercepted ROI, specifically, use a limited contrast adaptive histogram equalization algorithm (CLAHE).

S52. Perform Gabor four-direction filtering of 0°, 45°, 90°, and 135° on the ROI after the histogram equalization.

S6. Save the ROI after extracting a total of 3816 vein images of 636 categories;

S7, S7, using the method of one vs n when matching, using the direction of Zero Mean Normalized Cross-Correlation (ZNCC) to calculate the similarity score of the two ROI images, ranging from 0 to 1, the closer the value is to 1 The higher the degree of similarity is, the subject of the image with the highest similarity to the image to be matched is judged as the subject of the image to be matched.

The specific process is as follows:

S71. Firstly, classify all vein images.

S72. Select one as the image to be matched, divide the image to be matched and the candidate image into 8 rectangular areas, calculate the ZNCC value of the rectangular area corresponding to each pair of positions, and calculate the average ZNCC value of the two images as the similarity measure value, ZNCC The calculation requires the sum of the square differences of the two images, their respective grayscale averages, and standard deviations. Let the size of the two corresponding rectangular areas be (2n+1)×(2n+1), and the center coordinates are (u ₁ ,v ₁ ),(u ₂ ,v ₂ ), the sum of square differences is The average gray value is: The standard deviation is: ZNCC is: Take the average ZNCC value of 8 rectangular areas as the similarity measure standard. The ZNCC value ranges from 0 to 1. The closer the value is to 1, the higher the similarity is. The class of the image with the highest similarity to the image to be matched is judged as The category of the image to be matched.

S73. Calculate and sort the similarity between the image to be matched and the remaining images in the database, select the class of the image with the highest similarity as the predicted class, and compare it with the actual class. If they are equal, the prediction is correct, otherwise the prediction is wrong.

S74: The number of correct predictions during matching/total number of matches is the accuracy rate of the algorithm.

The image transformation process that ROI of the present invention extracts is represented in Fig. 3, and 3.1 represents the original image, and 3.2 represents the point set a extracted, and 3.3 represents the point set b extracted, and 3.4 represents the intersection of point set a and point set b, 3.5 represents the extended edge, 3.6 represents the vein image with the background gray value set to 0, 3.7 represents the vein image after correcting finger rotation, 3.8 represents the ROI image with histogram equalization, and 3.9 represents the use of four-directional Gabor filtering on the basis of 3.8 Imager ROI image.

As shown in Figure 4, the abscissa represents the false recognition rate, and the ordinate represents the false rejection rate, which changes with the set similarity threshold. When the prediction is correct but the similarity is less than the threshold, the false rejection rate is calculated, and the prediction is wrong but the similarity is false. When it is greater than the threshold, the false recognition rate is calculated. The 20*40, 40*80, and 80*160 in the lower right corner represent the resolution of the ROI after zooming in. The intersection of the dotted line and the solid line represents the equal error rate under different ROI resolutions. You can see It works best when the ROI is scaled to 40*80.

As shown in Figure 5, the abscissa represents the set similarity threshold, and the ordinate represents the accuracy rate. It can be seen that when the ROI resolution is 40*80, the accuracy rate is the highest when the similarity threshold is set to 0.68.

A finger vein recognition method and system proposed by the present invention, aimed at the shortcomings of the current finger vein recognition algorithm, adopts the ROI extraction method based on the finger contour, and at the same time increases the rotation correction operation of the finger, and enhances the finger deviation during image acquisition. The robustness of the rotation adopts the CLAHE histogram equalization to stretch the contrast of the image under different lighting conditions, which enhances the robustness to different lighting conditions. The four-directional Gabor filter is used to enhance the vein lines. Using the graph matching method of ZNCC, the ROI is divided into eight parts at the same time, and the similarity and average value are calculated respectively. The present invention is verified on the finger vein database disclosed by Shandong University, and the accuracy rate is about 98.7%. Since the vein image quality of this database is low, it can reach a higher accuracy under the condition of better vein image acquisition equipment. The accuracy rate can be widely used in identity authentication and recognition.

Embodiments of the present invention have been described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementations, and the above-mentioned specific implementations are only illustrative, rather than restrictive, and those of ordinary skill in the art will Under the enlightenment of the present invention, many forms can also be made without departing from the gist of the present invention and the protection scope of the claims, and these all belong to the protection of the present invention.

Claims (8)

1. A finger vein recognition method, characterized in that, comprising: S1. Determine the upper and lower edge point sets of the finger area in the original image, refine the edge to a pixel width, and place the finger area in the original image horizontally; S2. Select suitable points from the thinned edge point set to perform edge extension to obtain a real edge point set; S3. Correct the finger rotation, and set the gray value of the non-finger area to 0 according to the real edge point set; S4, the image processed through S3 is cropped, the width is selected as 0.73 ± 5% times of the original image, preferably 0.73 times, the bottom edge coordinates are selected when the upper edge is intercepted, and the uppermost edge coordinates are selected when the lower edge is intercepted. Highly intercepted to obtain the ROI of the vein region of interest; S5. Perform histogram equalization and Gabor filtering on the ROI of the vein ROI to obtain the ROI of the vein ROI after image enhancement, to be used for matching; S6. Using the steps of S1-S5 to process and extract vein ROI of multiple preset categories of vein images and save them; S7. The one vs n method is adopted for vein identification and matching, and the similarity score of each two vein ROI images is calculated using the zero-mean normalization method, and the vein ROI belonging to the same category is identified according to the similarity scores image. 2. A kind of finger vein recognition method according to claim 1, is characterized in that, in step S1, specifically comprises: S11. Obtain the edge point set a, if the difference between the gray value of the current pixel point and the gray value of the pixel point 2 coordinates above it is more than 33, then the pixel point is determined to be an edge point; S12. Obtain the edge point set b, and use the Sobel operator to calculate the gradient of the entire image. When the current pixel gradient exceeds the gradient value of two adjacent pixels in the gradient direction, the pixel is determined to be an edge point; S13. Perform an intersection operation on the edge point set a and the edge point set b to obtain the required edge point set, but only keep the first 15 pixels in the frequency of the abscissa, and refine the edge, that is, there are at most an upper edge point and a lower edge point; S14. The edge point set is represented in the image as a horizontal curve with uneven thickness, and thinning is performed on it, and the edge is thinned to a width of one pixel. 3. a kind of finger vein identification method according to claim 1, is characterized in that, in step S2, selects the point that a ordinate is closest to center respectively in the upper and lower edge point set as starting point extended edge, establishes this point (x , y), when extending to the left of the point, if the adjacent three coordinates (x-1, y-1), (x-1, y), (x-1, y+1) have a gray value 255, then set this adjacent point as the edge point, and continue to expand to the left. If there is no gray value of 255, then take the point with the largest gradient calculated in S1 by these three coordinates as the edge point, through continuous Extend to get the full finger outline. 4. A kind of finger vein recognition method according to claim 1, is characterized in that, in step S3, select four edge points (x1, y1), (x2, y1), (x3 , y2), (x4, y2), calculate the angle of finger rotation: Rotate the image to the horizontal according to the calculated rotation angle, and set the background gray value to 0. 5. A kind of finger vein recognition method according to claim 1, is characterized in that, in step S4, specifically comprises: S41. Use a rectangular window with a width of 50, move to the right from the middle coordinate of the vein, calculate the average gray value of the window every time you move a coordinate, and return the 5 window coordinates with the largest average gray value, and select the coordinates from these 5 The smallest one is used as the baseline of the ordinate, and the vein region of interest ROI of 0.73 times the width of the original image is intercepted to the left; S42. After the width interception is completed, perform height interception, select the bottom edge coordinates for the upper edge point set of the image, and select the uppermost edge coordinates for the lower edge point set to perform height interception to obtain the ROI of the vein ROI. 6. A kind of finger vein recognition method according to claim 1, is characterized in that, in step S5, specifically comprises: S51. Perform histogram equalization on the intercepted ROI, specifically, use a limited contrast adaptive histogram equalization algorithm; S52. Perform Gabor four-direction filtering of 0°, 45°, 90°, and 135° on the ROI after the histogram equalization. 7. A kind of finger vein recognition method according to claim 1, is characterized in that, in step S7, specifically comprises: S71. Firstly, classify all vein images; S72. Select one as the image to be matched, divide the image to be matched and the candidate image into 8 rectangular areas, calculate the ZNCC value of the rectangular area corresponding to each pair of positions, and calculate the average ZNCC value of the two images as the similarity measure value, ZNCC The calculation requires the sum of the square differences of the two images, their respective grayscale averages, and standard deviations. Let the size of the two corresponding rectangular areas be (2n+1)×(2n+1), and the center coordinates are (u ₁ ,v ₁ ),(u ₂ ,v ₂ ), the sum of square differences is The average gray value is: The standard deviation is: ZNCC is: The average ZNCC value of 8 rectangular areas is taken as the similarity measure standard. The ZNCC value ranges from 0 to 1. The closer the value is to 1, the higher the similarity is. The class of the image with the highest similarity to the image to be matched is judged as The category of the image to be matched; S73. Calculate and sort the similarity between the image to be matched and the remaining images in the database, select the class of the image with the largest similarity as the predicted class, and compare it with the actual class. If they are equal, the prediction is correct, otherwise the prediction is wrong; S74: The number of correct predictions during matching/total number of matches is the accuracy rate of the algorithm. 8. A finger vein recognition system, comprising: a processor and a storage device; the processor loads and executes instructions and data in the storage device for realizing any one of claims 1-7. A finger vein recognition method.