A method for recognizing a fingerprint image direction field

Abstract

The application discloses a fingerprint image direction field recognition method, comprising the following steps: S1, inputting a fingerprint image and dividing the input fingerprint image into regions; S2, calculating the gradient value matrix of two perpendicular directions of the center point of each divided part; S3, calculating the related angle trigonometric function value of the center point of each part in the original image according to the calculated gradient value matrix; and S4, calculating the direction field matrix of the center point of each part according to the obtained position value. The application realizes fingerprint recognition under the condition of limited hardware, and can endow a chip with low computing power and low storage space with fingerprint recognition capability. Meanwhile, the fingerprint image direction field recognition method provided by the application has high density of the obtained fingerprint image direction field, can more accurately and meticulously represent the direction field of the whole image, and thus improves the accuracy of fingerprint matching.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method for fingerprint image orientation field recognition. Background Technology

[0002] The complexity and uniqueness of fingerprints have led to their widespread use in unlocking devices. In daily life, fingerprint recognition technology has become an indispensable part of electronic products and digital management. However, noise inevitably appears when comparing and matching fingerprint images. A common approach is to extract the orientation field to assist in noise reduction. The fingerprint orientation field refers to assigning a representative orientation value to each small region of the fingerprint; concatenating all the orientation values ​​of the entire fingerprint yields an orientation matrix. Accurate extraction of the fingerprint orientation field directly affects whether a high-quality fingerprint can be obtained.

[0003] Methods for extracting fingerprint orientation fields mainly include gradient-based methods, model-based methods, filter-based methods, and convolutional neural network-based methods. Sherlock BG et al. proposed a simple mathematical model for calculating the local ridge orientation of a fingerprint. Kass M and Witkin A first proposed a gradient-based method, which is suitable for processing high-quality fingerprints but performs poorly for low-quality on-site fingerprints. Mei Y et al. also contributed to gradient-based methods. Compared to gradient-based methods, filter-based methods show better anti-interference capabilities, but have poor generalization ability, and the stitching of filtering results can lead to insufficient accuracy. Furthermore, they are slow and computationally intensive. Research on fingerprint recognition based on CNN networks has also yielded many results. Cao et al. used a method based on convolutional neural networks and template matching, which significantly improved fingerprint recognition performance compared to previous methods. Yao et al. replaced the traditional algorithm steps with convolutional neural networks, proposing an end-to-end fingerprint recognition network, FingerNet. Unlike fingerprint unlocking functions in common electronic products such as mobile phones, the hardware level of independent fingerprint unlocking devices such as electronic door locks is insufficient to achieve large storage capacity and complex calculations. Therefore, the CNN-based orientation field acquisition method is not applicable under low hardware conditions. Summary of the Invention

[0004] The purpose of this invention is to provide a fingerprint image orientation field recognition method to achieve fingerprint recognition function with low computing power and low storage chip requirements.

[0005] To address the aforementioned technical problems, this invention provides a fingerprint image orientation field recognition method, comprising the following steps:

[0006] S1. Input the fingerprint image and divide the input fingerprint image into regions;

[0007] S2. Calculate the gradient value matrix of the center point of each part in two mutually perpendicular directions.

[0008] S3. Calculate the trigonometric function matrix of the relevant angles of the center points of each part in the original figure based on the calculated gradient value matrix;

[0009] S4. Calculate the direction field matrix of the center point of each part based on the relevant angle trigonometric function matrix obtained above.

[0010] Furthermore, in step S1, the specific steps for dividing the region are as follows:

[0011] S11. Divide the fingerprint image into two equal halves, top and bottom;

[0012] S12. Divide both the upper and lower halves into four sections: left, right, upper center, and lower center.

[0013] S13. Divide the eight areas into multiple smaller grids.

[0014] Furthermore, before dividing the fingerprint image into smaller grids, the left, upper-middle, and right regions located in the upper half of the fingerprint image, as well as the left, lower-middle, and right regions located in the lower half of the fingerprint image, are filled with the required points around them using Gaussian filtering.

[0015] Furthermore, in step S2, the calculation of the gradient value matrices in two mutually perpendicular directions specifically includes the following steps:

[0016] S21. Calculate the gray value at the center point using linear interpolation.

[0017] S22. Perform convolution operation on the center point using the first Sobel operator to obtain matrix G. x ;

[0018] S23. Perform convolution operation on the center point using the second Sobel operator to obtain matrix G. y .

[0019] Furthermore, the first Sobel operator is The second Sobel operator is .

[0020] Furthermore, in step S3, the calculation of the corresponding trigonometric function matrix of the center point in the original image specifically includes the following steps:

[0021] S31, using G x and G y The two matrices are used to calculate G, which represents cos²θ and sin²θ respectively. d’ and G xy’Two matrices; θ is the angle perpendicular to the direction of the center point;

[0022] S32, G d’ and G xy’ By placing the two matrices into the corresponding positions in the original image, we obtain the trigonometric function matrix G of the relevant angles. d and G xy .

[0023] Furthermore, the G d’ and G xy’ The calculation formula is:

[0024] G d’ [i][j]=G x [i][j] 2 -G y [i][j] 2

[0025] G xy’ [i][j]=2×G x [i][j]×G y [i][j].

[0026] Furthermore, in step S4, the specific steps for calculating the center point direction field matrix are as follows:

[0027] S41. Divide the trigonometric function matrix of the relevant angles into multiple small grids;

[0028] S42. Apply a Gaussian kernel to the center point of each small grid and calculate the matrices representing cos2θ and sin2θ respectively based on the obtained values;

[0029] S43. Apply a Gaussian kernel to the obtained cos2θ and sin2θ matrices and assign the resulting values ​​to... The center point direction field matrix is ​​obtained through calculation.

[0030] Furthermore, in step S42, the matrix calculation formula for sin2θ is:

[0031] ;

[0032] The formula for calculating the matrix of cos2θ is:

[0033] .

[0034] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the fingerprint image orientation field recognition method as described above.

[0035] Compared with the prior art, the present invention has at least the following beneficial effects:

[0036] This invention enables fingerprint recognition under hardware constraints, providing fingerprint recognition capabilities to chips with low computing power and low storage space. Furthermore, the fingerprint image orientation field recognition method provided by this invention produces fingerprint images with high orientation field density, which can more accurately and meticulously represent the orientation field of the entire image, thereby improving the accuracy of fingerprint matching. Attached Figure Description

[0037] Figure 1 This is a flowchart of the fingerprint image orientation field recognition method in an embodiment of the present invention;

[0038] Figure 2 This is a fingerprint image region division diagram in an embodiment of the present invention;

[0039] Figure 3 This is a flowchart illustrating the specific operation of an image orientation field recognition method in a particular embodiment of the present invention.

[0040] Figure 4 This is a schematic diagram of angle θ in an embodiment of the present invention;

[0041] Figure 5 This is a diagram illustrating the fingerprint processing effect using existing technology.

[0042] Figure 6 This is a diagram illustrating the fingerprint processing effect of the present invention. Detailed Implementation

[0043] The fingerprint image orientation field recognition method of the present invention will be described in more detail below with reference to the schematic diagrams, which illustrate preferred embodiments of the present invention. It should be understood that those skilled in the art can modify the present invention described herein while still achieving the advantageous effects of the present invention. Therefore, the following description should be understood as being of broad knowledge to those skilled in the art and is not intended to limit the present invention.

[0044] The invention is described more specifically by way of example in the following paragraphs with reference to the accompanying drawings. The advantages and features of the invention will become clearer from the following description and claims. It should be noted that the drawings are in a very simplified form and use non-precise proportions, and are only used to facilitate and clarify the illustration of the embodiments of the invention.

[0045] The following are preferred embodiments of the fingerprint image orientation field recognition method to clearly illustrate the content of the present invention. It should be understood that the content of the present invention is not limited to the following embodiments, and other improvements made by conventional technical means by those skilled in the art are also within the scope of the present invention.

[0046] like Figure 1As shown, this embodiment of the invention proposes a fingerprint image orientation field recognition method, including:

[0047] S1. Input the fingerprint image and divide the input fingerprint image into regions;

[0048] Specifically, when dividing the fingerprint image into regions, the image is first divided into upper and lower halves. Then, each half is further divided into four regions: left, right, upper-middle, and lower-middle. The region division diagram is shown below. Figure 2 As shown, at this point, both the upper and lower halves contain four areas: left 1, right 2, upper center 3, and lower center 4. After the areas are divided, each area is further divided into multiple smaller grids.

[0049] Furthermore, before dividing the fingerprint image into small grids, Gaussian filtering is used to fill the surrounding points with the required data for the left 1, upper middle 3, and right 2 regions in the upper half of the fingerprint image, as well as the left 1, lower middle 4, and right 2 regions in the lower half of the fingerprint image.

[0050] S2. Calculate the gradient value matrix of the center point of each part in two mutually perpendicular directions.

[0051] Specifically, after dividing each region into small grids, since the center point of the selected grid does not actually exist in the original image matrix, it is necessary to obtain the grayscale value of the virtual point required for subsequent gradient calculation using linear interpolation before calculating the gradient. Then, the first Sobel operator is used. Perform a convolution operation on the center point of each small cell to obtain matrix G. x With the second Sobel operator Perform a convolution operation on the center point of each small cell to obtain matrix G. y .

[0052] S3. Calculate the trigonometric function matrix of the relevant angles of the center points of each part in the original figure based on the calculated gradient value matrix;

[0053] Specifically, two mutually perpendicular gradient value matrices G x and G y After the calculation is complete, use the formula: G d’ [i][j]=G x [i][j] 2 -G y [i][j] 2 G was calculated d’ Using the formula: G xy’ [i][j]=2×G x [i][j]×G y [i][j] calculates G xy’ G d’ and G xy’These are two matrices representing cos²θ and sin²θ, respectively, where θ is the angle perpendicular to the direction of the center point. -θ is the direction angle at the center point, where the direction at the midpoint of the direction field refers to the tangent direction of the ridge line, such as... Figure 4 The angle θ is illustrated schematically.

[0054] After obtaining G d’ and G xy’ After obtaining the value, place it at the corresponding position in the original image to obtain the relevant angle trigonometric function matrix G. d and G xy .

[0055] S4. Calculate the direction field matrix of the center point of each part based on the relevant angle trigonometric function matrix obtained above;

[0056] Specifically, G d and G xy It is worthwhile to get to G later. d and G xy Both matrices are divided into multiple small cells. Then, a Gaussian kernel is applied to the center point of each cell to filter the data, and the resulting values ​​are then analyzed using... and The matrices representing cos²θ and sin²θ are calculated, and then the calculated matrices are filtered using a Gaussian kernel, and then... The calculation yields an accurate direction field matrix representing the center point at the corresponding location.

[0057] In one specific embodiment, in conjunction with reference to Figure 3 The specific process is as follows:

[0058] Step 1: To reduce the space complexity of the calculation process and enable the algorithm to function as much as possible on chips with low storage capacity, the original 160×160 image matrix is ​​divided into upper and lower halves. Each half is further divided into four regions (left, right, upper-middle, and lower-middle) based on whether the surrounding points (i.e., the middle and surrounding parts of the image) need to be added during filtering. In other words, the original image is now divided into eight regions.

[0059] Step 2: Perform the following operations on each location in sequence:

[0060] (1) Divide the matrix into multiple small grids of size 2×2. For the center point of each small grid, use the Canny operator to calculate the gradient values ​​in the two perpendicular directions x and y and store them in matrix G. x G y middle:

[0061] In this step, an additional step is added for each of the six parts of the upper half matrix (left, top-middle, and right) and the lower half matrix (left, bottom-middle, and right): filling in the points needed for Gaussian filtering. Since the center points of the selected 2×2 grid do not actually exist in the original image matrix, linear interpolation is used to obtain the gray values ​​at these virtual points before calculating the gradient. Then, the Sobel operator is applied... Perform a convolution operation on the center point of each small cell to obtain matrix G. x Using the Sobel operator Perform a convolution operation on the center point of each small cell to obtain matrix G. y ;

[0062] (2) Using G x G y The two matrices yield G values ​​representing cos²θ and sin²θ (θ being the angle perpendicular to the direction of the center point), respectively. d’ G xy’ Two matrices:

[0063] G d’ [i][j]=G x [i][j] 2 -Gy[i][j] 2

[0064] G xy’ [i][j]=2*G x [i][j]*G y [i][j] ( )

[0065] After completing the above operations for each part, the resulting G will be... d’ and G xy’ The values ​​from the two matrices are placed into two matrices G of size 80×80. d and G xy The corresponding position.

[0066] Step 3: Perform the following operations on both Gd and Gxy matrices: Divide the matrix into 26×26 small cells of size 3×3, and filter the center point of each small cell using a 21×21 Gaussian kernel.

[0067] The values ​​obtained at the corresponding positions on Gd and Gxy are calculated using the following formula: , This yields two 26×26 matrices representing sin2θ and cos2θ, respectively.

[0068] Step 4: Filter the two matrices obtained above using a 9×9 Gaussian kernel, and then use... This yields a 26×26 matrix representing the accurate orientation field of the single point at the corresponding location.

[0069] Clearly, this method significantly reduces the computational and storage requirements during orientation field extraction by repeatedly performing operations on the center points of small blocks instead of on every point in the matrix. This is achieved by sequentially operating on the small blocks to fill in G. d G xy The two-matrix method replaces the direct operation on the entire image to obtain matrix G in one step. d G xy This reduces the number of intermediate variable matrices used in the calculation, significantly decreasing the algorithm's space complexity. Furthermore, by filtering the two matrices representing cos²θ and sin²θ and then recalculating cos²θ and sin²θ, and by operating on 2θ instead of θ during the process, errors are further reduced.

[0070] This method enables orientation field recognition of fingerprint images on a chip with a computing power of 168MHz and a storage space of 1M. Its spatial complexity is significantly reduced compared to methods that extract the trigonometric function matrix of the entire image without partitioning. Furthermore, it can be applied to chips with lower storage capacity and has high adaptability to chips under different conditions. The resulting... The single-point accurate orientation value matrix can meet the requirements of subsequent processing. Furthermore, multiple Gaussian filters were performed during the calculation process, resulting in a smoother orientation field that conforms to the characteristics of typical fingerprint images.

[0071] The implementation of the code incorporates assembly language, which further accelerates the actual calculation speed of the algorithm, reducing the running time by approximately 1.14%.

[0072] In terms of effectiveness, in conjunction with reference Figure 5 and Figure 6 As can be seen from the comparison, the fingerprint image orientation field obtained in the embodiments of the present invention ( Figure 5 and Figure 6 The dotted lines in the image are relatively denser, thus representing the orientation field of the entire image more accurately and in detail, thereby improving the accuracy of fingerprint matching.

[0073] In summary, this invention enables fingerprint recognition under hardware constraints, providing fingerprint recognition capabilities to chips with low computing power and limited storage space. Furthermore, the fingerprint image orientation field recognition method provided by this invention yields fingerprint images with high orientation field density, which can more accurately and meticulously represent the orientation field of the entire image, thereby improving the accuracy of fingerprint matching.

[0074] It should be understood that the embodiments described above are merely illustrative. The methods disclosed in the embodiments of the present invention can be implemented in various forms, such as by a device or other means.

[0075] For example, if the functions of this invention are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause the processor to execute all or part of the steps of the methods described in the various embodiments of this invention.

[0076] That is, those skilled in the art should understand that the embodiments of the present invention can be implemented in any of the following forms: a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects.

[0077] Based on this, embodiments of the present invention also provide a program product, which can be a storage medium such as a USB flash drive, external hard drive, ROM, RAM, magnetic disk, or optical disk. The storage medium can store a computer program, which, when run by a processor, executes the steps described in the foregoing method embodiments. The specific implementation and technical effects are similar and will not be repeated here.

[0078] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A fingerprint image orientation field recognition method, characterized in that, Includes the following steps: S1. Input the fingerprint image and divide the input fingerprint image into regions; S2. Calculate the gradient value matrix of the center point of each part in two mutually perpendicular directions. S3. Calculate the trigonometric function matrix of the relevant angles of the center points of each part in the original figure based on the calculated gradient value matrix; S4. Calculate the direction field matrix of the center point of each part based on the relevant angle trigonometric function matrix obtained above; In step S2, the calculation of the gradient value matrices in two mutually perpendicular directions specifically includes the following steps: S21. Calculate the gray value at the center point using linear interpolation. S22. Perform convolution operation on the center point using the first Sobel operator to obtain matrix G. x ; S23. Perform convolution operation on the center point using the second Sobel operator to obtain matrix G. y ; In step S3, the calculation of the trigonometric function matrix of the relevant angles of the center point in the original image specifically includes the following steps: S31, using G x and G y The two matrices are used to calculate G, which represents cos²θ and sin²θ respectively. d’ and G xy’ Two matrices; θ is the angle perpendicular to the direction of the center point; S32, G d’ and G xy’ By placing the two matrices into the corresponding positions in the original image, we obtain the trigonometric function matrix G of the relevant angles. d and G xy ; In step S4, the specific steps for calculating the center point direction field matrix are as follows: S41. Divide the trigonometric function matrix of the relevant angles into multiple small grids; S42. Apply a Gaussian kernel to the center point of each small grid and calculate the matrices representing cos2θ and sin2θ respectively based on the obtained values; S43. Apply a Gaussian kernel to the obtained cos2θ and sin2θ matrices and assign the resulting values ​​to... The center point direction field matrix is ​​obtained through calculation.

2. The fingerprint image orientation field recognition method as described in claim 1, characterized in that, In step S1, the specific steps for dividing the region are as follows: S11. Divide the fingerprint image into two equal halves, top and bottom; S12. Divide both the upper and lower halves into four sections according to the following method: left, right, upper center, and lower center; and S13. Divide the eight areas into multiple smaller grids.

3. The fingerprint image orientation field recognition method as described in claim 2, characterized in that, Before dividing the fingerprint image into smaller grids, the left, upper-middle, and right regions in the upper half of the fingerprint image, as well as the left, lower-middle, and right regions in the lower half of the fingerprint image, are filled with the required points around them using Gaussian filtering.

4. The fingerprint image orientation field recognition method as described in claim 1, characterized in that, The first Sobel operator is The second Sobel operator is .

5. The fingerprint image orientation field recognition method as described in claim 1, characterized in that, The G d’ and G xy’ The calculation formula is: G d’ [i][j]=G x [i][j] 2 -G y [i][j] 2 G xy’ [i][j]=2×G x [i][j]×G y [i][j]。 6. The fingerprint image orientation field recognition method as described in claim 1, characterized in that, In step S42, the matrix calculation formula for sin2θ is: ； The formula for calculating the matrix of cos2θ is: 。 7. A computer-readable storage medium storing a computer program thereon, characterized in that, The computer program is executed by the processor to perform the fingerprint image orientation field recognition method as described in any one of claims 1-6.

Images