A method for rapidly indexing a billion-level fingerprint library based on fingerprint feature distribution characteristics

Abstract

The application particularly relates to a method for rapidly indexing a billion-level fingerprint library based on fingerprint feature distribution characteristics. The method for rapidly indexing provided by the application comprises the following steps: step 1, fingerprint feature data discretization processing: fingerprint feature data rows are discretized by analyzing the adjacent distribution characteristics of fingerprint data; step 2, constructing a fingerprint feature similarity calculation model: the model is a matching matrix based on the adjacent distribution characteristics of fingerprint feature points; step 3, calculating the matching degree of sample fingerprints and template fingerprints: a matching score is calculated by establishing a fast matching vector of sample fingerprints; and step 4, selecting candidate fingerprints according to the matching degree. The method can rapidly index candidate fingerprints from a billion-level fingerprint library, reduces the consumption of computing resources, and effectively reduces the time required for fingerprint identification.

Description

Technical Field

[0001] This invention relates to the field of fingerprint recognition technology, specifically to a fast indexing method for a billion-level fingerprint database based on the distribution characteristics of fingerprint features. Background Technology

[0002] The skin texture of each person's fingers differs, manifested in the varying distribution, direction, and continuity of the ridges that make up the texture. This makes each person's fingerprint unique, thus enabling its use for identity verification. With years of technological advancements, fingerprint recognition, thanks to its rapid recognition speed and compact devices, has become the most widely used form of biometric authentication. It is extensively applied in mobile payments, access control systems, corporate attendance systems, and is also stored in ID cards to assist in identity verification.

[0003] Fingerprint recognition aims to extract features from collected fingerprints and match them with existing fingerprints in a database to achieve identity verification or identification. Fingerprint features can be divided into global features and local features. Global features include fingerprint pattern, core points, number of ridges, etc., serving as overall fingerprint features and used for coarse matching. Local features include ridge endpoints, bifurcation points, spurs, bridging points, etc., serving as detail features and used for precise identification. Among local features, endpoints and bifurcation points constitute a large portion of the local features and are considered the primary features for fingerprint recognition.

[0004] In the feature matching stage, the key to current fingerprint recognition technology lies in how to quickly and accurately match extracted features with fingerprints in the database. Existing fingerprint feature matching methods are mainly designed for one-to-one verification scenarios involving sample fingerprints and template fingerprints, and rarely address fast matching in one-to-many identification scenarios. Comparing sample fingerprints one by one with fingerprints in the database is more suitable for scenarios with a small number of fingerprints in the database. If the database stores a large number of fingerprints, comparing them one by one will generate significant computational costs and increase recognition time. Therefore, the speed of fingerprint identification in scenarios with massive amounts of data needs to be improved. Summary of the Invention

[0005] To address the problems of existing technologies, this invention provides a fast indexing method for a billion-level fingerprint database based on fingerprint feature distribution characteristics.

[0006] The fast indexing method provided by this invention includes the following steps:

[0007] Step 1: Discretization of fingerprint feature data: By analyzing the proximity distribution characteristics of fingerprint data, the fingerprint feature data is discretized.

[0008] Step 2: Construct a fingerprint feature similarity calculation model: The model is a matching matrix based on the proximity distribution features of fingerprint feature points;

[0009] Step 3: Calculate the matching degree between the sample fingerprint and the template fingerprint: Calculate the matching score by establishing a fast matching vector of the sample fingerprint;

[0010] Step 4: Select candidate fingerprints based on the degree of matching.

[0011] Furthermore, the fingerprint data mentioned in step 1 includes a set of fingerprint feature points of the samples to be matched. and the set of all template fingerprint feature points in the fingerprint database Where, p i and q n,j The fingerprint feature points are represented by (x, y, θ), where x, y, and θ correspond to the x-coordinate, y-coordinate, and orientation of the feature point in the image, respectively. m represents the number of feature points in the set of fingerprint feature points to be matched, and N represents the fingerprint feature points of the sample to be matched. n This represents the number of feature points of the nth fingerprint in the fingerprint database. There are a total of N template fingerprints in the fingerprint database.

[0012] Furthermore, in step 1, the discretization process for the fingerprint feature point sets P and Q is as follows:

[0013] S1. Quadrant division: Construct a three-dimensional rectangular coordinate system based on x, y and θ, and call the xOy plane the feature point position plane. Divide it into quadrants, with a total of A quadrants. Combine the x and y dimensions to form a three-dimensional space in the θ dimension, and divide it into intervals.

[0014] S2. Quadrant Interval Division: The intervals are divided according to the distance range from the point in the region to the origin and the direction angle. In xOy, each quadrant is divided into B fan-shaped ring regions according to the distance from the point to the origin. The direction angle is divided at unequal intervals. The basis for the division is to make the actual data fall evenly on each divided interval, and it is divided into Θ intervals.

[0015] S3. Feature extraction based on the nearest neighbor distribution of feature points: Traverse the feature points in the feature point set, and extract the traversed feature points p = (x p ,y p ,θ p ) of (x p ,y p ) is the origin of the coordinate system, and the direction is θ. p As the positive direction of the horizontal axis, another orthogonal direction is called the θ direction. Construct a three-dimensional rectangular coordinate system and divide the quadrants in the manner described in S1. Map the remaining feature points in the feature point set to this coordinate system. In each quadrant of xOy, if there are multiple feature points, only the feature point with the closest Euclidean distance to the origin is retained.

[0016] S4. Fingerprint Feature Discretization Processing: Within each quadrant of xOy, based on the Euclidean distance of the feature point from the origin, record its corresponding sector ring region. If no feature point exists within a quadrant, record it as belonging to the B-th sector ring region. Each quadrant of xOy has... There are several intervals, numbered 1, 2, ..., C. The interval information to which all feature points in each quadrant belong constitutes the neighborhood distribution feature of the fingerprint feature points. The neighborhood distribution feature of feature point p is defined as: L(p) = (l1(p), l2(p), ..., l... A (p)), where l a (p), a = 1, 2, ..., A represents the interval number of the neighboring feature points of the origin feature point p in the a-th quadrant. a (p)∈{1,2...,C}; In accordance with the above method, the fingerprint feature P to be processed and the fingerprint in the template fingerprint feature point set Q in the fingerprint database are discretized.

[0017] Furthermore, in step S2, the steps for constructing the matching matrix are as follows:

[0018] S5. Similarity calculation of fingerprint feature points p and q: Similarity is measured by analyzing their corresponding neighbor distribution features L(p) and L(q), with the neighboring points in the a-th quadrant belonging to the interval l. a (p), l a The distance (q) measures similarity. The closer the intervals are, the higher the similarity and the higher the matching score. If the interval does not exist in the quadrant or is too far away and falls into the Bth sector ring, the matching score is 0.

[0019] S6. Construction of the distribution feature matching matrix: The constructed matching matrix is ​​as follows Where the row direction and column direction represent the C of the feature points, respectively. A Given two neighbor feature distribution scenarios, for a feature point p of a fingerprint to be matched and a feature point q of a fingerprint in the fingerprint database, the matching matrix is ​​a grid data format where each column represents the neighbor feature distribution of different p, and each row represents the neighbor feature distribution of different q; M in the matching matrix j,i The matching score represents the matching score between points p and q when the neighboring feature distributions of feature point p are of the i-th type and the neighboring feature distributions of feature point q are of the j-th type. When i = j, the neighboring feature distributions of p and q are consistent, and the matching score is the highest.

[0020] Furthermore, the matching score in the a-th quadrant of S5 is calculated as follows:

[0021] Where s represents the scoring unit; then, the matching scores of fingerprint feature points p and q in all A quadrants are calculated as follows:

[0022] Furthermore, in step 3, the step of calculating the matching degree between the sample fingerprint and the template fingerprint is as follows:

[0023] S7. Establish the position index of the sample fingerprint P and all possible adjacency distributions of feature points about the distribution feature matching matrix: Perform cross-matching on the fingerprint point set discretization result obtained in step 1 and all possible adjacency distributions of feature points, based on the neighbor distribution features L(p) in the fingerprint point set discretization result. i Determine its column position information col(p) in the distribution feature matching matrix M. i Thus, the sample fingerprint set P can be represented as a set of column indices with respect to the distribution feature matching matrix M, expressed as: Among them, col(p) i )∈{1,2,...,C A The template fingerprint point set Qn is represented as a set of row information about the distribution feature matching matrix M, expressed as: Where row(q) n,j L(q) represents the neighborhood distribution characteristic. n,j The corresponding row position in the distribution feature matching matrix, and row(q) n,j )∈{1,2,...,C A}, further calculate the set of row information corresponding to the distribution features of all template fingerprints in the fingerprint database;

[0024] S8. Calculate the sample fingerprint P and a template fingerprint Q. n The fast matching score; calculate the j-th row of the distribution feature matrix M corresponding to col(p1), col(p2), ..., col(p m The maximum value in column E is represented as: E j (p)=max{M j,col(p1) M j,col(p2) ,...,M j,col(pm)}, thus constructing a fast matching vector E(P) for feature point P, expressed as: For the template fingerprint feature point set Q n Based on the row position information of the feature points in the feature matching matrix The element at the corresponding position is indexed in the fast matching vector E(P) of the feature point set P. P,Q n The quick matching score is the sum of the elements at all corresponding positions;

[0025] S9, Sample fingerprint P and all template fingerprints Q in the fingerprint database nMatch score calculation: The sample fingerprint is matched with the fingerprints in the fingerprint database one by one to calculate the matching score, and the matching score set with the fingerprints in the fingerprint database is obtained.

[0026] Furthermore, step 4 specifically involves:

[0027] Based on the matching scores between the sample fingerprint P from step 3 and all fingerprints in the fingerprint feature database, the fingerprint with the highest matching score is selected. The fingerprint features corresponding to each item are used as candidate fingerprints for subsequent precise fingerprint feature matching.

[0028] Furthermore, it is characterized in that, The value range is 4 to 20.

[0029] The present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described fast indexing method.

[0030] The present invention also provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the steps of the above-described fast indexing method.

[0031] The beneficial effects of this invention are as follows:

[0032] (1) By analyzing the proximity distribution characteristics of fingerprint points, the fingerprint feature point set is discretized to reduce data complexity, improve the robustness to fingerprint feature anomalies, facilitate the storage of fingerprint feature calculation results, and thus avoid redundant calculations.

[0033] (2) By constructing a matching matrix based on the proximity distribution features of fingerprint feature points, a fast index of the matching degree of fingerprint feature points can be achieved;

[0034] (3) By establishing a fast matching vector for sample fingerprints, the matching speed with fingerprint features in the fingerprint database is further improved; thus, a fast candidate fingerprint selection method is constructed before precise matching of fingerprint features, which can quickly index candidate fingerprints from a database of one billion fingerprints, reducing the consumption of computing resources and effectively reducing the time required for fingerprint identification. Attached Figure Description

[0035] Figure 1 This is a flowchart of the method of the present invention;

[0036] Figure 2 This is a schematic diagram of the fingerprint distribution feature matching matrix of the present invention;

[0037] Figure 3 This is a schematic diagram of the matching matrix of the present invention;

[0038] Figure 4This is a schematic diagram illustrating the matching score calculation of the present invention. Detailed Implementation

[0039] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] This embodiment provides a fast indexing method for a billion-level fingerprint database based on fingerprint feature distribution characteristics, such as... Figure 1 As shown, it includes the following steps:

[0041] Step 1: Discretization of fingerprint feature data;

[0042] Step 2: Construct a fingerprint feature similarity calculation model;

[0043] Step 3: Calculate the degree of matching between the sample fingerprint and the template fingerprint;

[0044] Step 4: Select candidate fingerprints based on the degree of matching.

[0045] In step 1, the fingerprint data includes: the fingerprint feature point set, which includes the fingerprint feature point set of the samples to be matched. and the set of all template fingerprint feature points in the fingerprint database Where, p i and q n,j The fingerprint feature points are represented by (x, y, θ), where x, y, and θ correspond to the x-coordinate, y-coordinate, and orientation of the feature point in the image, respectively. m represents the number of feature points in the set of fingerprint feature points to be matched, and N represents the fingerprint feature points of the sample to be matched. n This represents the number of feature points of the nth fingerprint in the fingerprint database. There are N template fingerprints in the fingerprint database, where N is on the order of billions of data points.

[0046] In step 1, the discretization process for the fingerprint feature point sets P and Q is as follows:

[0047] 1.1 Quadrant Division: A three-dimensional rectangular coordinate system is constructed based on x, y, and θ, and the xOy plane is called the feature point position plane. Quadrants are divided into A quadrants. In this embodiment, A = 4. The three-dimensional space is formed in the θ dimension by combining the x and y dimensions, and intervals are formed by dividing it.

[0048] 1.2 Quadrant Interval Division: Intervals are divided based on the distance range from points within the region to the origin and the directional angle. In the xOy plane, each quadrant is divided into B equal fan-shaped annular regions based on the distance from the point to the origin. For the directional angle, the division is done at unequal intervals, aiming to distribute the actual data as evenly as possible across the divided intervals and to minimize the number of intervals, resulting in Θ intervals. In this implementation, for the xOy plane, each quadrant is divided into B = 10 fan-shaped annular regions based on the distance from the point to the origin. For the directional dimension, after statistical analysis, the interval [0, π] is divided at unequal intervals symmetrical with respect to π / 2, Θ = 3, resulting in [0, π / 6], (π / 6, 5π / 6], and (5π / 6, π]. The open and closed intervals during the division process can be adjusted automatically without affecting the statistical regularity.

[0049] Given a minimum and maximum distance difference of ρ pixels between points in each sector ring region and the origin, and dividing the sector ring regions into equal segments with distances of ρ, the range of distances from points within the i-th sector ring region to the origin is represented as:

[0050]

[0051] 1.3 Feature extraction based on the nearest neighbor distribution of feature points: Traverse the feature points in the feature point set, and extract the traversed feature points p = (x p ,y p ,θ p ) of (x p ,y p ) is the origin of the coordinate system, and the direction is θ. p As the positive direction of the horizontal axis, another orthogonal direction is called the θ direction. A three-dimensional Cartesian coordinate system is constructed, and quadrants are divided as described in 1.1. The remaining feature points in the feature point set are mapped to this coordinate system. In each quadrant on xoy, if multiple feature points exist, only the feature point with the closest Euclidean distance to the origin is retained, such as... Figure 2 As shown.

[0052] 1.4 Fingerprint Feature Discretization: Within each quadrant of xOy, the sector ring region to which the feature point belongs is recorded based on its Euclidean distance from the origin. If no feature point exists within a quadrant, it is recorded as belonging to the B-th sector ring region. The θ dimension of a non-existent point is also unknown; therefore, the θ dimension of the B-th sector ring region is not divided. For each quadrant of xOy, the 1st, 2nd, ..., B-1th sector ring regions have Θ divisions corresponding to the θ dimension, while the B-th sector ring region has only one division. Therefore, each quadrant has... There are several intervals, and these buckets are numbered 1, 2, ..., C according to a certain pattern. The interval information to which all feature points in each quadrant belong constitutes the neighborhood distribution feature of the fingerprint feature points. Therefore, the neighborhood distribution feature of each fingerprint feature point has C. AOne possible representation. The neighborhood distribution feature of feature point p is defined as: L(p) = (l1(p), l2(p), ..., l... A (p)), where l a (p), a = 1, 2, ..., A represents the interval number of the neighboring feature points of the origin feature point p in the a-th quadrant. a (p)∈{1,2...,C}; Following the above method, the fingerprint feature P to be processed is discretized to obtain... The nth fingerprint Q from the template fingerprint feature point set Q in the fingerprint database. n Discretize into

[0053] In step 2, the fingerprint feature similarity calculation model is constructed as follows:

[0054] 2.1 Similarity calculation of fingerprint feature points p and q: Similarity is measured by analyzing their corresponding neighbor distribution features L(p) and L(q), with the neighboring points in the a-th quadrant belonging to the interval l. a (p), l a The distance (q) measures similarity; the closer the intervals are, the higher the similarity, and thus the higher the matching score. If the interval does not exist in that quadrant or is too far away and falls into the B-th sector ring, the matching score is 0. The matching score in the a-th quadrant is calculated as follows: Where s represents the scoring unit.

[0055] Then, the matching scores for feature points p and q across all A quadrants are calculated as follows:

[0056]

[0057] 2.2 Construction of Distribution Feature Matching Matrix:

[0058] Based on all possible neighbor feature distribution representations and similarity calculation rules from step 1, a matching matrix is ​​constructed to achieve the goal of quickly retrieving the matching scores of the two feature distributions. The constructed matching matrix is ​​as follows: Where the row direction and column direction represent the C of the feature points, respectively. A Given two neighbor feature distribution scenarios, for a feature point p of a fingerprint to be matched and a feature point q of a fingerprint in the fingerprint database, the matching matrix is ​​a grid data format where each column represents the neighbor feature distribution of different p, and each row represents the neighbor feature distribution of different q; M in the matching matrix j,i Let represent the matching score between feature point p and feature point q when the neighbor feature distributions are of the i-th type and feature point q is of the j-th type. When i = j, the neighbor feature distributions of p and q are identical, resulting in the highest matching score, s. Taking k = 4 and c = 3 as an example, the matching matrix representation is as follows: Figure 3 As shown.

[0059] In step 3, the specific steps for calculating the matching degree between the sample fingerprint and the template fingerprint are as follows:

[0060] 3.1 Establish the position index of the sample fingerprint P and the possible adjacency distributions of all feature points with respect to the distribution feature matching matrix:

[0061] 3.1.1 Discretization results of the fingerprint point set obtained in step 1 Perform cross-matching with all possible adjacent distributions of feature points, based on... The neighbor distribution characteristics L(p) i Determine its column position information col(p) in the distribution feature matching matrix M. i Thus, the sample fingerprint set P can be represented as a set of column indices with respect to the distribution feature matching matrix M, expressed as: Among them, col(p) i L(p) represents the neighborhood distribution characteristic. i The corresponding column position in the distribution feature matching matrix, and col(p) i )∈{1,2,...,C A}

[0062] 3.1.2 Similarly, the template fingerprint point set Q n The set of row information about the distribution feature matching matrix M is represented as follows:

[0063] Where row(q) n,j L(q) represents the neighborhood distribution characteristic. n,j The corresponding row position in the distribution feature matching matrix, and row(q) n,j )∈{1,2,...,C A Furthermore, the set of row information corresponding to the distribution features of all template fingerprints in the fingerprint database can be calculated.

[0064] 3.2 Sample fingerprint P and a template fingerprint Q n Match score calculation:

[0065] In the feature matching matrix M, a feature point p from the sample fingerprint point set... i The corresponding column and q j The value at the position determined by the corresponding row is the feature point p. i With feature point q j The matching score is expressed as Therefore, the feature point p can be obtained. i With template feature point set Q n The set of matching scores for all feature points in the set, where feature point pi With template feature point set Q n The best matching score is the maximum value in the set, expressed as: Therefore, the sample fingerprint P and the template fingerprint Q n The matching score is expressed as:

[0066] 3.3 Sample fingerprint P and a template fingerprint Q n Quick match score;

[0067] Given a set of fingerprint feature points P to be identified, calculate the best matching score between it and all possible fingerprint template feature points. According to the definition above, this requires calculating the feature points q in the entire fingerprint database. n,j And calculate the corresponding score(P,Q) n When the fingerprint database contains a large number of fingerprints, the computational load for this step becomes enormous. Therefore, the following approximate gradual change algorithm is adopted.

[0068] Calculate its columns col(p1), col(p2), ..., col(p) in the distribution feature matrix M. m Then, calculate the scores corresponding to all possible interval combinations. This calculation only requires performing a simple matrix maximization and indexing operation. Specifically, calculate the scores corresponding to the j-th row of the distribution feature matrix M, where col(p1), col(p2), ..., col(p...). m The maximum value in the column is represented as: Thus, a fast matching vector E(P) for feature point P is constructed, expressed as: For the template fingerprint feature point set Q n Based on the row position information of the feature points in the feature matching matrix The element at the corresponding position is indexed in the fast matching vector E(P) of the feature point set P. like Figure 4 As shown, then P,Q n The matching score is the sum of the elements at all corresponding positions, expressed as:

[0069] 3.4 Calculation of the matching score between the sample fingerprint P and all template fingerprints Q in the fingerprint database:

[0070] The sample fingerprint is matched one by one with the fingerprints in the fingerprint database, and the matching score is calculated to obtain the set of matching scores with the fingerprints in the fingerprint database, which is represented as:

[0071] Step 4: Select candidate fingerprints based on the degree of matching.

[0072] Based on the matching scores (score(P,Q)) between the sample fingerprint P from step 3 and all fingerprints in the fingerprint feature database, the fingerprints with the highest matching scores are selected. The fingerprint features corresponding to each item are used as candidate fingerprints for subsequent precise fingerprint feature matching. The value range is 4 to 20.

Claims

1. A fast indexing method for a billion-level fingerprint database based on fingerprint feature distribution characteristics, characterized in that, The method includes the following steps: Step 1: Discretization of fingerprint feature data: By analyzing the proximity distribution characteristics of fingerprint data, the fingerprint feature data is discretized; the fingerprint data includes the set of fingerprint feature points of the samples to be matched. and the set of template fingerprint feature points in the fingerprint database. ;in, and Representing fingerprint feature points, each fingerprint feature point is... It means that, among them 、 and These correspond to the x-coordinate, y-coordinate, and orientation of the feature point in the image, respectively. This indicates the number of feature points in the set of fingerprint feature points of the sample to be matched. Indicates the number of fingerprints in the database The number of feature points in a fingerprint, and the total number of fingerprint database entries. A template fingerprint; a set of fingerprint feature points and The discretization process involves the following steps: S1, quadrant division: based on... 、 and Construct a three-dimensional Cartesian coordinate system and... The plane is called the feature point location plane, and it is divided into quadrants, totaling... Each quadrant, combined with , Dimensions A three-dimensional space is formed in terms of dimensions, and it is divided into intervals; S2, Quadrant interval division: the intervals are divided according to the distance range from points within the region to the origin and the direction angle. Above, each quadrant is divided into equal parts according to the distance from the point to the origin. A sector-shaped annular region; divided at unequal intervals with varying directional angles, the division is based on ensuring that the actual data falls evenly across each interval, thus dividing the region into several sectors. Intervals; S3, Feature point neighborhood distribution feature extraction: Traverse the feature points in the feature point set, and extract the traversed feature points... of As the origin of the coordinate system, direction As the positive direction of the horizontal axis, the other orthogonal direction is called... Construct a three-dimensional Cartesian coordinate system and divide the quadrants according to the method in S1; map the remaining feature points in the feature point set to this coordinate system. In each quadrant of the fingerprint, if multiple feature points exist, only the feature point with the closest Euclidean distance to the origin is retained; S4, fingerprint feature discretization processing: In Within each quadrant, the sector region to which the feature point belongs is recorded based on its Euclidean distance from the origin. If no feature point exists in a quadrant, it is recorded as belonging to the [missing quadrant name]. A fan-shaped ring region, Each quadrant has There are several intervals, numbered as follows: The interval information to which all feature points belong in all quadrants constitutes the neighborhood distribution characteristics of fingerprint feature points; feature points The neighborhood distribution characteristic is defined as: ,in , Represents the origin feature points In the The numbering of the interval to which the adjacent feature points within the quadrant belong. Following the above method, the fingerprint features to be processed... P And the template fingerprint feature point set in the fingerprint database The fingerprints in the image are discretized. Step 2: Construct a fingerprint feature similarity calculation model: The model is a matching matrix based on the neighborhood distribution features of fingerprint feature points. The steps for constructing the matching matrix are: S5, fingerprint feature points... Similarity calculation: by analyzing the corresponding neighborhood distribution characteristics , To measure similarity, in the first The intervals of neighboring points within each quadrant , The distance between intervals measures similarity; the closer the intervals, the higher the similarity, and thus the higher the matching score. If an interval does not exist in that quadrant or is too far away, it falls into the second quadrant. Within each sector-shaped ring, the matching score is 0; S6, Construction of the distribution feature matching matrix: The constructed matching matrix is Where the row direction and column direction represent the feature points respectively. The distribution of neighboring features, for the feature points of the fingerprint to be matched. and the feature points of a fingerprint in the fingerprint database The matching matrix uses each column to represent a distinct feature. The distribution of neighboring features, each row representing a different The grid data form is formed by the distribution of neighboring features; in the matching matrix Representing feature points The neighborhood feature distribution is the first Distribution of species, characteristic points The neighborhood feature distribution is the first Under what distribution conditions, The matching score of two points; when hour, The nearest neighbor features are consistent, resulting in the highest matching score; Step 3: Calculate the matching degree between the sample fingerprint and the template fingerprint. The steps are as follows: S7: Establish the sample fingerprint. Position index of the fingerprint point set and all possible adjacent distributions of feature points with respect to the distribution feature matching matrix: Perform cross-matching between the discretized fingerprint point set obtained in step 1 and all possible adjacent distributions of feature points, based on the neighbor distribution features in the discretized fingerprint point set. Determine its matching matrix in the distribution features The corresponding column position information Thus, the sample fingerprint point set Represented as a matching matrix with respect to distribution features The set of column indexes is represented as: ,in, ; Set template fingerprint points Represented as a matching matrix with respect to distribution features The set of row information is represented as: ,in, Indicates neighborhood distribution characteristics The corresponding row position in the distribution feature matching matrix, and Further calculate the set of row information corresponding to the distribution features of all template fingerprints in the fingerprint database; S8, calculate the sample fingerprints. With a certain template fingerprint Fast matching score; calculate distribution feature matrix The row correspondence The maximum value in the column is represented as: Thus constructing information about feature points Fast matching vector , is represented as: For template fingerprint feature point set Based on the row position information of the feature points in the feature matching matrix In the feature point set fast matching E ( P The element at the corresponding index position in ) ; , The fast matching score is the sum of all corresponding position elements; S9, sample fingerprint. Compared with all template fingerprints in the fingerprint database Match score calculation: The sample fingerprint is matched with the fingerprints in the fingerprint database one by one to calculate the matching score, and the matching score set with the fingerprints in the fingerprint database is obtained. Step 4: Select candidate fingerprints based on the degree of matching.

2. The fast indexing method according to claim 1, characterized in that, S5 in the The matching score within each quadrant is calculated as follows: ,in, Indicates the scoring unit; then in all In each quadrant, fingerprint feature points The matching score calculation result is as follows: .

3. The fast indexing method according to claim 2, characterized in that, Step 4 specifically involves... Based on the sample fingerprints from step 3 The fingerprints are compared with all fingerprints in the fingerprint feature database, and the fingerprints with the highest matching scores are selected. The fingerprint features corresponding to each item are used as candidate fingerprints for subsequent precise fingerprint feature matching.

4. The fast indexing method according to claim 3, characterized in that, The value range is 4 to 20.

5. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-4.

6. A computer-readable storage medium for storing computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the method according to any one of claims 1-4.

Images