A fingerprint orientation field reconstruction matching method

Abstract

The present invention provides a fingerprint direction field reconstruction and matching method. The fingerprint feature point information is known, the direction field is reconstructed according to the feature point information, and a feature point quality evaluation system is established. The feature points and the reconstructed direction field are combined to perform fingerprint matching. In the process of reconstructing the fingerprint direction field, the feature points and the direction field are mutually optimized and iterated; secondly, a local direction field of the direction area block around the feature point is established, which can be used for local direction field matching calculation. The method of the present invention increases the dimension of calculating the similarity score in the fingerprint matching process, improves the precision and accuracy of fingerprint matching, and also improves the comprehensive performance of the matching algorithm.

Description

Technical Field

[0001] The invention belongs to the field of image processing and relates to a fingerprint direction field reconstruction matching method. Background Art

[0002] Fingerprints are unique to the human body. The complexity of fingerprints is sufficient to meet the needs of identification, and fingerprints are easy to collect. Therefore, fingerprint recognition has become one of the most popular, convenient and reliable personal identity authentication technologies. The fingerprint recognition process mainly includes fingerprint image acquisition, fingerprint image preprocessing, fingerprint feature extraction, feature matching and matching result output. Fingerprint image acquisition is to enter the fingerprint ridge distribution information through relevant equipment; fingerprint image preprocessing is to perform certain processing on the collected fingerprint image, such as image segmentation, fingerprint enhancement, image refinement, image binarization, etc.; fingerprint feature extraction is to extract fingerprint feature information from the preprocessed image; fingerprint feature matching is to determine whether two fingerprints are homologous based on the extracted feature information. The fingerprint direction field describes the directional pattern information of the fingerprint ridges and valleys, which can be obtained in the fingerprint preprocessing step. As a global and reliable feature of fingerprints, fingerprint direction field calculation plays a vital role in fingerprint image enhancement, fingerprint singular point detection, fingerprint feature point extraction, etc.

[0003] The feature point information of fingerprints is a compressed representation of fingerprint image information and is the most widely used feature in fingerprint matching. In some specific fields, such as public security criminal investigation, banking and insurance services, in order to save storage space, fingerprint images are generally not directly stored, only feature point templates are stored, and then fingerprint matching is performed based on the position and direction information of the fingerprint feature points. However, this matching method can use fewer information dimensions, and some fingerprint feature points also contain some interference noise points, so fingerprint matching may be difficult. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a fingerprint direction field reconstruction and matching method. Based on the known fingerprint feature point information, the direction field is reconstructed according to the feature point information, and a feature point quality evaluation system is established. The feature points and the reconstructed direction field are combined for fingerprint matching. In the process of reconstructing the fingerprint direction field, the feature points and the direction field are mutually optimized and iterated; secondly, a local direction field of the direction area block around the feature point is established, which can be used for local direction field matching calculation. The method of the present invention increases the dimension of scoring in the fingerprint matching process and improves the accuracy of fingerprint matching.

[0005] To achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0006] The present invention discloses a fingerprint direction field reconstruction and matching method. The fingerprint feature point information is known, the direction field is reconstructed according to the feature point information, and a feature point quality evaluation system is established. The fingerprint matching is performed by combining the feature points and the reconstructed direction field. The specific steps are as follows:

[0007] S1: Divide the fingerprint image into evenly distributed blocks, and the direction in each block is consistent;

[0008] S2: Radiate to the surrounding area with each fingerprint feature point as the center in a certain way to reconstruct the entire fingerprint image direction field;

[0009] S3: Establish a quality evaluation system, combine the direction and position information of the fingerprint feature point and its surrounding feature points, evaluate the quality of each fingerprint feature point, obtain the comprehensive quality score of each feature point, and classify the credibility of the feature point according to the comprehensive quality score of each feature point;

[0010] S4: Fingerprint matching calculation is performed by combining feature points, comprehensive quality scores of feature points and reconstructed direction fields.

[0011] In the above technical solution, the fingerprint feature point is an endpoint or a fork point obtained according to the fingerprint image; the fingerprint feature point information includes the position information of the fingerprint feature point and the actual direction information of the fingerprint feature point, that is, the actual direction along the ridge direction with the point as the starting point. In this technical solution, the reconstruction of the fingerprint direction field mainly comes from the information of the fingerprint feature point. Each fingerprint feature point information contains the direction of the point (0°-360°), and the direction field information is (-90°-90°) or (0°-180°). According to the fingerprint characteristics, each feature point represents the direction field of the area around the point by default. Therefore, with the area block to which each feature point belongs as the center, radiating to the surrounding areas in a certain way, the entire fingerprint image direction field can be reconstructed.

[0012] The quality evaluation of fingerprint feature points is a common method for fingerprint matching. This technical solution establishes a quality evaluation system to conduct a detailed analysis and evaluation of the surrounding area of ​​each feature point. The evaluation parameters include but are not limited to the coordinate position, type, direction, distortion and other information of each feature point. These indicators are combined with the quality evaluation system to calculate the comprehensive quality score of the feature point, and the feature points are classified according to the comprehensive quality score of each feature point for subsequent fingerprint matching calculations. The comprehensive quality score of the feature point can be used to correct the weight of the matching score accumulation process of each feature point, that is, the weight of the fingerprint matching can be designed according to the comprehensive quality score of each feature point, and then used for fingerprint matching.

[0013] Furthermore, in the above technical solution, each area block includes a direction vector, and the direction of the area block where the feature point is located is the direction of the feature point, and then radiates to the surrounding area in a certain manner with the area block to which each feature point belongs as the center, specifically including: for any feature point, the size of the direction vector module of the area block where the feature point is located is set to 1, and with the position of the area block where the feature point is located as the center, the directions of the surrounding area blocks of the feature point are the same as the direction of the area block where the feature point is located, and the size of the direction vector module of the surrounding area blocks changes based on a specific functional relationship with distance as a variable; the specific function includes but is not limited to: normal distribution function, Gaussian function or inverse curve function.

[0014] In the above technical solution, the direction vector of the area block where the non-feature point is located is the sum of the direction vectors generated by the radiation of each feature point according to the weight, and the weight is related to the comprehensive quality score of the feature point; the direction of the area block where the non-feature point is located is the direction corresponding to its direction vector.

[0015] As a technical solution, by evaluating the quality and classifying the credibility of fingerprint feature points, for high-credibility feature points, a local direction field of the surrounding directional area block is established, and local direction field matching is performed to increase the scoring dimension and improve the comprehensive matching evaluation ability. For high-credibility feature points, a local direction field of the surrounding directional area block is established and introduced into the local direction field matching as the weight of the local area matching, such as star-match matching, triangle matching, MCC matching and other fingerprint local matching methods, to increase the dimension and accuracy of local matching and improve the comprehensive evaluation ability. Star-match matching (star map matching) is to match the stars in the observed image with the reference stars in the star catalog to establish the corresponding relationship between the CCD star image and the star. Through the star-mtach algorithm, that is, the star map matching algorithm (such as the polygon angular distance matching algorithm, the triangle matching algorithm, the grid algorithm and the algorithm using the star-to-axis image template, etc.), the theoretical star map and the measured star map can be quickly matched, and it is also widely used in fingerprint recognition. MCC is a cylindrical coding algorithm based on a three-dimensional data structure. It constructs cylindrical coding based on the distance and angle of the details. The MCC algorithm is also widely used in the field of fingerprint matching. In addition, this technical solution can also expand the matching dimension for global matching methods such as line pair matching (i.e., obtaining the best rotation and translation information of two fingerprint images, performing the best overlap processing, calculating the overlap of each feature point information after overlap, examining the position coordinates and direction information, etc.), that is, after overlap, in addition to the information matching of the feature points, the overall direction field of the regional block to which the feature points belong can also be locally matched, which can improve the matching accuracy.

[0016] As a technical solution, according to the credibility level of the fingerprint feature points, the feature points with the lowest credibility level are removed, and the above steps S2 and S3 are repeated to iteratively reconstruct the fingerprint direction field. In this technical solution, fingerprint matching is mainly based on high-credibility fingerprint feature points, but a part of the extracted fingerprint feature point information may be noise points, and the number of noise points will directly affect the matching accuracy. By evaluating the quality of the fingerprint feature points and classifying the credibility, information such as noise points can be preliminarily distinguished. For the feature points with the lowest credibility level, they are determined to be noise points and deleted, which can greatly improve the matching accuracy. Repeat the above steps S2 and S3 to iteratively reconstruct the fingerprint direction field, and the matching results can be optimized through iterative reconstruction.

[0017] Furthermore, in the above technical solution, multi-order surface fitting is performed on the reconstructed direction field to achieve a smooth and coherent effect of the direction field.

[0018] As a technical solution, a low proportion of matching weight is assigned to the feature points with the lowest credibility level, and the above steps S2 and S3 are repeated, and then the optimization can be continuously performed progressively.

[0019] As a technical solution, the reconstruction process of the fingerprint direction field also includes the acquisition of singular points, which can be introduced into the matching mechanism to increase the matching dimension and improve the matching accuracy. The singular points of fingerprints include core points and triangulation points. The core point is the highest point on the curved ridge of the fingerprint pattern, and the triangulation point is the center of the triangle area formed when three different directional flows meet together. Based on the uniqueness of fingerprints, many fingerprint classification, fingerprint matching and recognition algorithms are based on the extraction of fingerprint singular points. Obtaining fingerprint singular points and their positions is of great significance to fingerprint matching. For example, if one fingerprint has a singular point and the other does not, the matching weight can be reduced; for example, if the singular point positions of the two fingerprints are the same, the matching weight can be increased. In addition, a large number of literatures have introduced algorithms for extracting and detecting fingerprint singular points, such as methods based on morphological analysis, filtering, calculation of Poincare index values, neural networks, genetic algorithms, directional histograms, Gaussian-Hermite matrices, complex filters, etc. Preferably, in the present technical solution, a Markov random field model is constructed according to the directional region block matrix and the directional probability field, and singular points are detected based on the constructed model.

[0020] Beneficial effects of the present invention: The present invention provides a fingerprint direction field reconstruction and matching method. The fingerprint feature point information is known, the direction field is reconstructed according to the feature point information, and a feature point quality evaluation system is established. The feature points and the reconstructed direction field are combined to perform fingerprint matching. In the process of reconstructing the fingerprint direction field, the feature points and the direction field are mutually optimized and iterated; secondly, a local direction field of the directional area block around the feature point is established, which can be used for local direction field matching calculation. The method of the present invention introduces some other information, such as the comprehensive quality score of the feature points, pre-processes these feature points and reconstructs the fingerprint direction field, which is of great significance to the matching of fingerprints. Secondly, the method of the present invention increases the dimension of calculating the similarity score in the fingerprint matching process, improves the precision and accuracy of fingerprint matching, and also improves the overall performance of the matching algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A schematic diagram of a method flow chart of an embodiment of the present invention;

[0022] Figure 2 A schematic diagram of the fingerprint feature point information diagram and the direction of the surrounding area in one embodiment of the present invention;

[0023] Figure 3 An image after multi-order surface fitting is performed according to an embodiment of the present invention;

[0024] Figure 4 A schematic diagram showing a comparison between a direction field reconstructed based on fingerprint feature point information and a direction field directly obtained from an image in one embodiment of the present invention;

[0025] Figure 5 It is a schematic diagram of comparison between a direction field reconstructed based on fingerprint feature point information and a direction field directly obtained from an image in another embodiment of the present invention;

[0026] Figure 6 It is a schematic diagram comparing the direction field reconstructed based on fingerprint feature point information and the direction field directly obtained from the image in yet another embodiment of the present invention. DETAILED DESCRIPTION

[0027] The specific embodiments of the present invention are described below in conjunction with the accompanying drawings, but it should be understood by those skilled in the art that the following embodiments are only for illustration and not for limiting the scope of the present invention. It should be understood by those skilled in the art that the following embodiments may be modified without departing from the scope and spirit of the present invention. The scope of protection of the present invention is defined by the appended claims.

[0028] This embodiment provides a fingerprint direction field reconstruction matching method, the method flow chart is as follows Figure 1 As shown, specifically including:

[0029] S1: Divide the fingerprint image into evenly distributed blocks, and the direction in each block is consistent;

[0030] S2: Radiate to the surrounding area with each fingerprint feature point as the center in a certain way to reconstruct the entire fingerprint image direction field;

[0031] S3: Establish a quality evaluation system, combine the direction and position information of the fingerprint feature point and its surrounding feature points, evaluate the quality of each fingerprint feature point, obtain the comprehensive quality score of each feature point, and classify the credibility of the feature point according to the comprehensive quality score of each feature point;

[0032] S4: Fingerprint matching calculation is performed by combining feature points, comprehensive quality scores of feature points and reconstructed direction fields.

[0033] like Figure 2 As shown, Figure 2 (a) is a fingerprint feature point information diagram of this embodiment, which clearly shows the direction information of each fingerprint feature point and the evenly distributed area blocks obtained by its division; Figure 2 (b) is a schematic diagram of the direction of the area around each fingerprint feature point after radiating from the area block to the surrounding area. It can be seen that the fingerprint feature point information affects the direction of the surrounding area, and the direction of the area block where the non-feature point is located is the superposition of the directions generated by the radiation of all feature points. In addition, Figure 2 (b) There is a singular point in the circle that is inconsistent with the feature direction information of other area blocks. The directions of the three feature points in the circled area are different from its fitting trajectory. The size of the area block can be set according to needs. Different sizes of area blocks will affect the number and method of subsequent loop iterations.

[0034] Preferably, a fifth-order surface fitting is performed on the reconstructed direction field to achieve a smooth and coherent direction field, such as Figure 3 shown.

[0035] Preferably, the reconstruction process of the fingerprint direction field also includes obtaining singular points, introducing the singular points into the matching mechanism, and increasing the matching dimension.

[0036] In some specific fields, such as public security criminal investigation, banking and insurance services, fingerprint matching is generally based on the position information and direction information of the feature points, rather than the original fingerprint image directly obtained, that is, the direction field cannot be directly obtained based on the fingerprint image. Therefore, the fingerprint feature points are pre-processed through the above technical solution, and then the fingerprint direction field is reconstructed, which can greatly improve the accuracy of fingerprint matching.

[0037] In one embodiment of the present invention, Figure 4 As shown, Figure 4 (a) represents the original fingerprint image. Figure 4(b) represents the direction field image obtained from the direct image. Figure 4 (c) represents fingerprint feature points and texture information, Figure 4 (d) represents simple feature point information, Figure 4 (e) indicates that the Figure 4 (d) The direction field reconstructed by the feature point information in the above method steps. As described in the above method steps, the fingerprint direction field is reconstructed based on the provided fingerprint feature points and texture information, the fingerprint feature points are evaluated for quality, and the credibility level is classified, and the fingerprint matching calculation is performed by combining the feature points, the comprehensive quality scores of the feature points and the reconstructed direction field; according to the credibility level of the fingerprint feature points, the feature points with the lowest credibility level are removed, and the fingerprint direction field is iteratively reconstructed; multi-order surface fitting is performed on the reconstructed direction field to achieve a smooth and coherent effect of the direction field, based on Figure 4 The direction field reconstructed from the feature point information in (d) is as follows Figure 4 (e) Figure 4 (e) and the direction field obtained directly from the fingerprint image, namely Figure 4 (b) is compared and the difference is not big. It can be seen that in this technical solution, by reconstructing the fingerprint direction field and introducing the matching weight of the feature point quality score, the matching interference of the noise point is weakened or removed, or the matching of the feature point with high credibility is enhanced, so as to improve the matching accuracy.

[0038] In another embodiment of the present invention, Figure 5 As shown, Figure 5 (a) represents the original fingerprint image. Figure 5 (b) represents the direction field image obtained from the direct image. Figure 5 (c) represents fingerprint feature points and texture information, Figure 5 (d) represents simple fingerprint feature point information, Figure 5 (e) indicates that the Figure 5 (d) The direction field reconstructed from the fingerprint feature point information. This embodiment introduces local matching of local feature information. The circled area in the figure is the local direction field. For high-confidence feature points, the local direction field of the surrounding direction area blocks is established to perform local direction field matching to avoid erroneous matching caused by the inconsistency of the local direction area blocks while the feature point position and direction information meet the matching requirements, thereby further improving the matching accuracy. Figure 5 (e) with Figure 5 (b) By comparison, it can also be seen that the introduction of local direction field matching through this method can further improve the matching accuracy. In addition, through local direction field matching, the matching quality is higher.

[0039] In another embodiment of the present invention, Figure 6 As shown, Figure 6 (a) represents the original fingerprint image. Figure 6 (b) represents the direction field image obtained from the direct image. Figure 6 (c) represents fingerprint feature points and texture information, Figure 6 (d) represents simple feature point information, Figure 6 (e) indicates that the Figure 6 (d) The direction field reconstructed from the feature point information. Figure 6 In (d), the circle area changes continuously, indicating that the reliability of the quality point here is relatively high. The direction field formed by the characteristic points in this area encloses a direction change area, and the singular point can be determined; Figure 6 In (e), after the direction field is reconstructed in the box area, it is detected that the change in this area is relatively serious and the change is discontinuous on all sides. It is judged that there may be noise points or low-quality points. The main reason for the distortion of the direction field caused by the directional deviation around the area block is the feature points in the box. Therefore, a low-proportion matching weight processing is performed on this feature point, which can then overlap with the actual direction field to obtain a higher matching accuracy.

[0040] Preferably, in the above embodiment, the credibility weight of a specific feature point can be further set in detail in the algorithm model, and the quality evaluation of a certain feature point needs to be evaluated with reference to the continuity of the surrounding directional field.

[0041] Preferably, in the feature point coincidence matching, similarity score evaluation dimensions, local matching and other methods are added based on the local direction field distribution of the feature points and the type and position of the singular points to optimize the matching effect and improve the matching accuracy.

Claims

1. A fingerprint direction field reconstruction matching method, characterized in that: Given the fingerprint feature point information, the direction field is reconstructed based on the feature point information, and a feature point quality evaluation system is established. Fingerprint matching is performed by combining the feature points and the reconstructed direction field. The specific process is as follows: S1: Divide the fingerprint image into evenly distributed blocks, and the direction in each block is consistent; S2: radiate to the surrounding area in a certain way with the area block to which each fingerprint feature point belongs as the center, so as to reconstruct the direction field of the entire fingerprint image; each area block includes a direction vector, the direction of the area block where the feature point is located is the direction of the feature point, and radiate to the surrounding area in a certain way with the area block to which each feature point belongs as the center, specifically including: for any feature point, the magnitude of the direction vector modulus of the area block where the feature point is located is set to 1, and the direction of each area block around the feature point is the same as the direction of the area block where the feature point is located with the position of the area block where the feature point is located as the center, and the magnitude of the direction vector modulus of each area block around the feature point changes based on a specific function relationship with the distance as a variable, and the specific function includes: normal distribution function, Gaussian function or inverse curve function; S3: Establish a quality evaluation system, combine the direction and position information of the fingerprint feature point and its surrounding feature points, evaluate the quality of each fingerprint feature point, obtain the comprehensive quality score of each feature point, and classify the credibility of the feature point according to the comprehensive quality score of each feature point; S4: Perform fingerprint matching calculation based on the feature points, their comprehensive quality scores and the reconstructed direction field. According to the credibility level of the fingerprint feature points, remove the feature points with the lowest credibility level, repeat steps S2 and S3, and iteratively reconstruct the fingerprint direction field; or, assign a low proportion of matching weights to the feature points with the lowest credibility level, repeat steps S2 and S3, and then continuously optimize progressively.

2. According to claim 1, a fingerprint direction field reconstruction matching method is characterized in that: The direction vector of the area block where the non-feature point is located is the sum of the direction vectors generated by the radiation of each feature point according to the weight, and the weight is related to the comprehensive quality score of the feature point; the direction of the area block where the non-feature point is located is the direction corresponding to its direction vector.

3. The fingerprint direction field reconstruction matching method according to claim 1, characterized in that: For high-confidence feature points, the local direction field of the surrounding directional area blocks is established, and local direction field matching is performed to increase the scoring dimension and improve the comprehensive matching evaluation capability.

4. The fingerprint direction field reconstruction matching method according to claim 1, characterized in that: Multi-order surface fitting is performed on the reconstructed direction field to achieve a smooth and coherent direction field.

5. The fingerprint direction field reconstruction matching method according to claim 1, characterized in that: The reconstruction process of the fingerprint direction field also includes the acquisition of singular points, which are introduced into the matching mechanism to increase the matching dimension.

6. A fingerprint direction field reconstruction matching method according to claim 5, characterized in that: The acquisition of the singular point includes constructing a Markov random field model according to the directional region block matrix and the directional probability field, and obtaining the singular point based on the constructed model.

Images