A fast fingerprint search method based on quadtree structure encoding

Abstract

This invention provides a fast fingerprint search method based on quad-branch structure encoding, including a registration process and a search process. The registration process records all fingerprints to be registered in a database for searching, including extracting minutiae of the fingerprints to be registered; constructing a quad-branch structure for each minutiae; encoding each quad-branch structure; and storing the encoded string and related information in the database. The search process searches for the fingerprint most similar to the fingerprint to be searched in the registered fingerprint database, including extracting the encoding of the fingerprint to be searched, searching the encoding in the database, calculating the similarity between the fingerprint to be searched and the fingerprints in the database based on statistical information, and sorting them in descending order of similarity. By sorting the fingerprints to be matched in the fingerprint database according to their matching probability through the search, the number of comparisons required to match the corresponding fingerprint is significantly reduced, greatly improving fingerprint recognition efficiency.

Description

Technical Field

[0001] This invention relates to the field of fingerprint recognition technology, and more specifically, to a fast fingerprint search method based on quad-branch structure encoding. Background Technology

[0002] With the widespread adoption of fingerprint recognition technology in modern society, the size of fingerprint template databases is also growing rapidly, posing a significant challenge to fingerprint recognition systems. On the one hand, to ensure the security and accuracy of fingerprint recognition systems, the speed of fingerprint recognition is limited to a certain extent; on the other hand, the large-scale fingerprint database places demands on the speed of fingerprint recognition systems.

[0003] To balance the trade-off between accuracy and speed in fingerprint recognition, a common approach is to reduce the range of fingerprints to be compared, thus decreasing the number of comparisons. Two main approaches exist: fingerprint classification and fingerprint search. Fingerprint classification algorithms categorize fingerprints into arched, arc-shaped, and ring-shaped types, but the distribution of each category is uneven, and it doesn't effectively reduce the comparison range with large datasets. Fingerprint search methods are generally divided into those based on global features and those based on minutiae features. Global feature-based methods are often not robust enough to fingerprint deformation and displacement, while minutiae-based methods are more stable. Common minutiae-based search methods often suffer from complex or excessive feature structures. For example, the feature structure used in correlation methods based on MCC cylinder structures is a multi-layered complex structure. In correlation methods based on Delaunay's triangular model and those based on feature point connection pairs, the number of features often equals the total number of permutations of all minutiae features because each minutiae needs to establish connections with as many other minutiae as possible. Summary of the Invention

[0004] The present invention aims to at least solve one of the technical problems in the prior art, namely, the large amount of computation and long time consumption of 1:N matching of fingerprints under large-scale datasets, the difficulty in simultaneously meeting the requirements of fingerprint recognition accuracy and speed, the lack of robustness of global feature-based methods to fingerprint deformation and displacement, and the complexity or excessive number of feature structures in feature search methods based on minutiae feature points.

[0005] Therefore, this invention provides a fast fingerprint search method based on quad-branch structure encoding.

[0006] This invention provides a fast fingerprint search method based on quad-branch structure encoding, including a registration process and a search process;

[0007] The registration process records all fingerprints to be registered in the database for searching, including extracting the minutiae of the fingerprints to be registered; constructing a quad-branch structure for each minutiae; encoding each quad-branch structure; and storing the encoded string and related information in the database.

[0008] The search process involves searching for the fingerprint most similar to the fingerprint to be searched in the registered fingerprint database. This includes extracting the code of the fingerprint to be searched, searching the code in the database, calculating the similarity between the fingerprint to be searched and the fingerprints in the database based on statistical information, and sorting them in descending order of similarity to obtain the search results.

[0009] According to the above-described fingerprint fast search method based on quad-branch structure encoding of the present invention, it may further have the following additional technical features:

[0010] In the above technical solution, the quadrilateral structure includes one central detail feature point, four surrounding detail feature points, and four edges. Taking any detail feature point as the central detail feature point, with the coordinates of the central detail feature point as the center, the direction of the central detail feature point as the x-axis, and the direction perpendicular to the x-axis as the y-axis, the local planar space is divided into four quadrants. The four surrounding detail feature points are each taken as the points in these four quadrants closest to the central detail feature point. Connecting the surrounding detail feature points to the central detail feature point yields four edges. Therefore, the quadrilateral structure can be represented as:

[0011] T i =(E ia E ib E ic E id ) T

[0012] Where E ia E ib E ic E id These are the four sides of a quadrilateral structure.

[0013] In the above technical solution, each edge of the quadrilateral structure is represented by a three-dimensional vector:

[0014]

[0015] Among them, E ij d represents the edge between the central detail feature point and the surrounding detail feature points. ij θ represents the Euclidean distance between the central detail feature point and the surrounding detail feature points. ij This represents the angle difference between the direction of the central detail feature point and the direction of the surrounding detail feature points. This represents the difference between the angle of the line connecting the coordinates of the central detail feature point and the coordinates of the surrounding detail feature point and the angle of the direction of the central detail feature point.

[0016] In the above technical solution, the angle difference θ between the direction of the central detail feature point and the direction surrounding the detail feature point is... ij The calculation method is as follows:

[0017] θ ij =diff(θ) i θ j )

[0018] Where, θ i θ represents the direction of the central detail feature point. j This represents the direction surrounding the detailed feature point.

[0019] In the above technical solution, the angle between the line connecting the coordinates of the central detail feature point and the surrounding detail feature point differs from the angle of the direction of the central detail feature point. The calculation method is as follows:

[0020]

[0021] Where, x i With y i The x-coordinate represents the coordinate position of the central detail feature point. j With y j Represents the coordinates of the surrounding detail feature points.

[0022] In the above technical solution, 3D-GeoHash is used to process each edge vector E. ij Encode it into a comparable and easily stored string.

[0023] In the above technical solution, the 3D-GeoHash includes the following steps:

[0024] a. Divide the three-dimensional vector space into two parts alternately according to the three coordinate axes. If the component value is in the first half of the coordinate axis during each division, the binary bit of that part is 0, otherwise it is 1. Each division is called a round of division. Each round of division makes the precision range of the encoding smaller. When the precision reaches the required range, the division stops and the binary code H is obtained.

[0025] b. Encode the quadrant into a 2-bit binary number and concatenate it with H;

[0026] c. Encode the binary sequence into a string using BASE-8 or BASE-16.

[0027] In the above technical solution, the search process includes the following steps:

[0028] Establish a four-way structure hit table and a fingerprint hit table. The four-way structure hit table is responsible for counting the number of times a certain four-way structure of a certain fingerprint is hit; the fingerprint hit table is responsible for counting the number of times a certain fingerprint is hit.

[0029] Extract the detailed feature point template of the fingerprint to be searched and generate a set of quad-branch structures. Encode each quad-branch structure into a quad string encoding group using 3D-GeoHash.

[0030] Iterate through all string codes of the quad structure and search in the database. If the code exists in the database, increment the count of a quad structure for a fingerprint by 1 in the quad structure hit table.

[0031] The four-way structure hit table is statistically analyzed. If the hit count is greater than a certain threshold, the corresponding fingerprint count in the fingerprint hit table is incremented by 1.

[0032] The final fingerprint similarity score is calculated by combining the number of quadrangular structures in the fingerprint to be searched, the number of quadrangular structures in the matched fingerprints, and the fingerprint matching table, and then output in descending order of score.

[0033] In any of the above technical solutions, the relevant information includes the fingerprint image sequence number and the four-fork structure sequence number.

[0034] In any of the above technical solutions, the fingerprint to be registered includes several minutiae feature points, each of which is a five-element feature vector, and the features include coordinate position, orientation, category, and quality confidence.

[0035] In summary, due to the adoption of the above-mentioned technical features, the beneficial effects of the present invention are:

[0036] In terms of feature structure, this method proposes a feature description method based on a four-way structure of detail feature points, which constructs features only in the local neighborhood of detail feature points, thus reducing both the number of features and the feature complexity. In terms of feature storage, this method proposes an extended Geo-Hash, which encodes features into string format, making it compatible with most databases, easy to access, and highly versatile.

[0037] After integrating this method into the fingerprint recognition system, the fingerprints to be matched in the fingerprint database are sorted according to the matching probability by searching, which greatly reduces the number of comparisons required to match the corresponding fingerprint. In the case of 1:N comparison of a large-scale fingerprint database, it can greatly reduce the time required to find the correct match and greatly improve the fingerprint recognition efficiency.

[0038] This invention encodes features into a single string, making it compatible with common databases on the market. It eliminates the need to consider additional features storage and query processes, which is beneficial for the storage and retrieval of large-scale data.

[0039] The quadrilateral structure proposed in this invention involves only 5 points in the local neighborhood of the feature point, resulting in a small overall computational load. Furthermore, the intermediate computational results generated can be reused in most fingerprint 1:1 matching algorithms, avoiding redundant calculations and effectively improving the overall computational efficiency of the fingerprint recognition system.

[0040] Additional aspects and advantages of the invention will become apparent in the following description or may be learned by practice of the invention. Attached Figure Description

[0041] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0042] Figure 1 This is a schematic diagram of the quad-branch structure in a fingerprint fast search method based on quad-branch structure encoding according to an embodiment of the present invention;

[0043] Figure 2 This is a schematic diagram of the fingerprint registration process in a fingerprint fast search method based on quad-branch structure encoding according to an embodiment of the present invention;

[0044] Figure 3 This is a schematic diagram of the fingerprint search process in a fingerprint fast search method based on quad-branch structure encoding according to an embodiment of the present invention;

[0045] Figure 4 This is a schematic diagram of fingerprint hit counting in a fingerprint fast search method based on quad-branch structure encoding according to an embodiment of the present invention. Detailed Implementation

[0046] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0047] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0048] The following reference Figures 1 to 4 This describes a fast fingerprint search method based on quad-branch structure encoding, provided by some embodiments of the present invention.

[0049] Some embodiments of this application provide a fast fingerprint search method based on quad-branch structure encoding.

[0050] like Figures 1 to 4As shown, the first embodiment of the present invention proposes a fast fingerprint search method based on quad-branch structure encoding, including a registration process and a search process;

[0051] The registration process records all fingerprints to be registered in the database for searching, including extracting the minutiae of the fingerprints to be registered; constructing a quad-branch structure for each minutiae; encoding each quad-branch structure; and storing the encoded string and related information in the database.

[0052] The search process involves searching for the fingerprint most similar to the fingerprint to be searched in the registered fingerprint database. This includes extracting the code of the fingerprint to be searched, searching the code in the database, calculating the similarity between the fingerprint to be searched and the fingerprints in the database based on statistical information, sorting them in descending order of similarity, obtaining the search results, and matching the corresponding fingerprint.

[0053] The second embodiment of this invention proposes a fast fingerprint search method based on a quad-branch structure encoding, and builds upon the first embodiment, as follows: Figure 1 As shown, the fingerprint to be registered includes several minutiae feature points, each of which is a five-element feature vector, and the features include coordinate position, orientation, category, and quality confidence.

[0054] Let M = {m1, m2, ..., m n} represents the set of minutiae templates extracted from the fingerprint F to be registered, conforming to the ISO / IEC 19794-2 standard. Each minutiae m i It is a five-element eigenvector, m i ={x i ,y i ,θ i ,t i ,q i}, where x i With y i Representing detail feature point m i The coordinate position, θ i t represents its pointing direction. i Representing its category, q i This represents the quality reliability of the detailed feature point.

[0055] The four-pronged structure includes one central detail feature point, four surrounding detail feature points, and four edges. Taking any detail feature point as the central detail feature point, with the coordinates of the central detail feature point as the center, the direction of the central detail feature point as the positive x-axis, and the direction perpendicular to the x-axis as the positive y-axis, the local planar space can be divided into four quadrants in a counter-clockwise direction. The four surrounding detail feature points are each taken as the points in these four quadrants closest to the central detail feature point. Connecting the surrounding detail feature points to the central detail feature point yields the four edges. The fingerprint F is then calculated using the formula m. i For the central detailed feature point, m a ,m b ,m c ,m d A four-pronged T structure surrounding detailed feature points i It can be represented by 4 vectors:

[0056] T i =(E ia E ib E ic E id ) T

[0057] Where E ia E ib E ic E id These are the four sides of a quadrilateral structure.

[0058] Each edge in the quadrilateral structure is represented by a three-dimensional vector:

[0059]

[0060] Among them, E ij d represents the edge between the central detail feature point and the surrounding detail feature points. ij θ represents the Euclidean distance between the central detail feature point and the surrounding detail feature points. ij This represents the angle difference between the direction of the central detail feature point and the direction of the surrounding detail feature points. This represents the difference between the angle of the line connecting the coordinates of the central detail feature point and the coordinates of the surrounding detail feature point and the angle of the direction of the central detail feature point.

[0061] The angle difference θ between the direction of the central detail feature point and the direction of the surrounding detail feature points ij The calculation method is as follows:

[0062] θ ij =diff(θ) i θ j )

[0063] Where, θ iθ represents the direction of the central detail feature point. j This represents the direction surrounding the detailed feature point.

[0064] The difference between the angle of the line connecting the coordinates of the central detail feature point and the surrounding detail feature points and the angle of the direction of the central detail feature point. The calculation method is as follows:

[0065]

[0066] Where, x i With y i The x-coordinate represents the coordinate position of the central detail feature point. j With y j Represents the coordinates of the surrounding detail feature points.

[0067]

[0068] Minute details that are too close together are most likely caused by sensor errors and detection algorithm errors, while points that are too far apart usually have large angular differences due to fingerprint deformation. To make the algorithm more robust, minute details that are too close or too far apart should be excluded when selecting surrounding minute details.

[0069] The third embodiment of the present invention proposes a fast fingerprint search method based on quad-branch structure encoding, and, based on any of the above embodiments, such as... Figures 1 to 4 As shown, for the four-pronged structure T i =(E ia E ib E ic E id ) T 3D-GeoHash is used to process each edge vector E ij Encode it into a comparable and easily stored string.

[0070] The common spatial indexing encoding technique Geo-Hash is only used for two-dimensional latitude and longitude encoding and is not suitable for the three-dimensional edge vectors designed in this method. This invention extends Geo-Hash to 3D-GeoHash to realize the function of encoding three-dimensional vectors and incorporating quadrant information into the encoding.

[0071] The 3D-GeoHash includes the following steps:

[0072] a. Divide the three-dimensional vector space into two parts alternately according to the three coordinate axes. If the component value is in the first half of the coordinate axis during each division, the binary bit of that part is 0, otherwise it is 1. Each division is called a round of division. Each round of division makes the precision range of the encoding smaller. When the precision reaches the required range, the division stops and the binary code H is obtained.

[0073] b. Encode the quadrant into a 2-bit binary number and concatenate it with H;

[0074] c. Encode the binary sequence into a string using BASE-8 or BASE-16.

[0075] Four-fork structure T i After the above spatial encoding conversion, it can be represented as a quadruple string encoding group. If the string encodings of two edge vectors are the same, then their corresponding edge vectors are in the same quadrant in the quadruple structure and have similar properties.

[0076] like Figure 2 As shown, for a fingerprint image F to be registered i Extract its detail feature point templates and generate a quadrilateral structure set T = {T1, T2, ..., T}. n}, for each quadrilateral structure T j The fingerprint image is encoded into a quadruple string using 3D-GeoHash. Finally, the string encoding, fingerprint image sequence number, quadruple structure sequence number, and other relevant information are inserted into the database.

[0077] The fourth embodiment of this invention proposes a fast fingerprint search method based on quad-branch structure encoding, and, based on any of the above embodiments, such as... Figure 3 As shown, the search process includes the following steps:

[0078] Establish a four-way structure hit table and a fingerprint hit table. The four-way structure hit table is responsible for counting the number of times a certain four-way structure of a certain fingerprint is hit; the fingerprint hit table is responsible for counting the number of times a certain fingerprint is hit.

[0079] Extract the detailed feature point template of the fingerprint to be searched and generate a set of quad-branch structures. Encode each quad-branch structure into a quad string encoding group using 3D-GeoHash.

[0080] Iterate through all string codes of the quad structure and search in the database. If the code exists in the database, increment the count of a quad structure for a fingerprint by 1 in the quad structure hit table.

[0081] The four-way fingerprint hit table is analyzed. If the hit count is greater than a certain threshold, the corresponding fingerprint count in the fingerprint hit table is incremented by 1. The process is as follows: Figure 4 As shown;

[0082] The final fingerprint similarity score is calculated by combining the number of quadrangular structures in the fingerprint to be searched, the number of quadrangular structures in the matched fingerprints, and the fingerprint matching table, and then output in descending order of score.

[0083] In this specification, the illustrative expressions of the terms used do not necessarily refer to the same embodiments or examples. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0084] Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention shall be included within the scope of protection of this invention.

Claims

1. A method for fast search of fingerprints based on encoding of quadtree structure, characterized by, The method comprises a registration process and a search process; The registration process registers all to-be-registered fingerprints into a database for searching, including extracting detail feature points of the to-be-registered fingerprints; constructing a quadtree structure for each detail feature point; and encoding each quadtree structure; The string and related information obtained by encoding are stored in the database; The search process searches the registered fingerprints in the database for the most similar fingerprint to the to-be-searched fingerprint, including extracting the code of the to-be-searched fingerprint, searching the code in the database, calculating the similarity between the to-be-searched fingerprint and the fingerprints in the database according to the statistical information, and arranging the fingerprints in descending order of similarity; The quadtree structure comprises one central detail feature point, four surrounding detail feature points and four edges; any detail feature point is taken as the central detail feature point, the coordinate of the central detail feature point is taken as the center, the direction of the central detail feature point is taken as the x-axis direction, and the direction perpendicular to the x-axis is taken as the y-axis direction, so as to divide the local plane space into four quadrants; the four surrounding detail feature points are the points closest to the central detail feature point in the four quadrants, and the four edges are obtained by connecting the surrounding detail feature points and the central detail feature point, so that the quadtree structure can be expressed as: wherein respectively are 4 edges in a quad structure; Each edge in the quadtree structure is represented by a three-dimensional vector: wherein, represents the edge between the center detail feature point and the surrounding detail feature point, represents the Euclidean distance between the center detail feature point and the surrounding detail feature point, represents the angle difference between the center detail feature point direction and the surrounding detail feature point direction, represents the angle difference between the angle of the line connecting the center detail feature point and the surrounding detail feature point coordinate points and the angle of the line connecting the center detail feature point and the surrounding detail feature point coordinate points, represents the angle difference between the angle of the line connecting the center detail feature point and the surrounding detail feature point coordinate points and the angle of the line connecting the center detail feature point and the surrounding detail feature point coordinate points, Difference between the angle of the center detail feature point direction and the angle of the surrounding detail feature point direction The calculation method is: wherein, representing a center detail feature point direction, representing a surrounding detail feature point direction; the angle of the line connecting the center detail feature point and the surrounding detail feature point coordinate point and the angle of the center detail feature point the difference of the angle The calculation method is: wherein, with a coordinate position representing a center detail feature point, with a coordinate position representing a surrounding detail feature point.

2. The method according to claim 1, wherein, Each edge vector is encoded as a comparable and easily storable string using 3D-GeoHash. Encoding as a comparable and easily storable string.

3. The method according to claim 2, wherein, The 3D-GeoHash comprises the following steps: a. The three-dimensional vector space is alternately divided into two parts along the three coordinate axes; if the component value is in the first half of the coordinate axis, the binary bit is 0, otherwise, the binary bit is 1; three divisions are called one round of division, and each round of division makes the accuracy range of the code smaller; when the accuracy reaches the required range, the division is stopped, and a binary code H is obtained; b. The quadrant code is encoded into a two-bit binary number and concatenated with H; c. The binary sequence is encoded into a string by BASE-8 or BASE-16.

4. The method according to claim 3, wherein, The search process comprises the following steps: A quadtree structure hit table and a fingerprint hit table are established, the quadtree structure hit table is responsible for counting the number of hits of a quadtree structure of a fingerprint, and the fingerprint hit table is responsible for counting the number of hits of a fingerprint; A detail feature point template of the to-be-searched fingerprint is extracted, and a quadtree structure set is generated, each quadtree structure is encoded into a four-element string code group by 3D-GeoHash; All string codes of the quadtree structure are traversed, and a search is performed in the database; if the code exists in the database, the count of a quadtree structure of a certain fingerprint in the quadtree structure hit table is increased by 1; The quadtree structure hit table is counted, and if the hit count is greater than a certain threshold, the count of the corresponding fingerprint in the fingerprint hit table is increased by 1; The final similarity score of the fingerprint is calculated by combining the number of quadtree structures in the to-be-searched fingerprint, the number of hit quadtree structures of the fingerprint and the fingerprint hit table, and the score is output in descending order.

5. The method according to any one of claims 1 to 4, wherein, The related information comprises a fingerprint image serial number and a quadtree structure serial number.

6. The method according to any one of claims 1 to 4, wherein, The to-be-registered fingerprint comprises a plurality of detail feature points, each detail feature point is a five-element feature vector, and the features include coordinate position, direction, category and quality reliability.

Images