Financial data identity authentication method based on fingerprint recognition
By analyzing the directionality of fingerprint grayscale graphs and binary graphs, quantifying and correcting glitch dispute parameters, the glitch problem of OPTA algorithm in financial data identity authentication is solved, and the efficient and accurate fingerprint recognition is achieved.
Patent Information
- Application Number
- CN202410727329.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-06
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-06-06
AI Technical Summary
Existing fingerprint refinement algorithms such as OPTA algorithm have glitch problems in financial data identity authentication, resulting in inaccurate feature extraction and insufficient recognition accuracy, which cannot meet the requirements of efficiency.
By obtaining the fingerprint grayscale map and the single pixel point width, refine the fingerprint binary map, analyze the binary directionality and grayscale directionality of the pixel point, quantify the initial glitch dispute parameters, and correct the directional difference coefficient to finally obtain the glitch-free 指纹图像进行身份认证。
It effectively eliminates the glitches in fingerprint refinement images, improves the accuracy and efficiency of fingerprint recognition, and ensures the efficiency and accuracy of financial data identity authentication.
Smart Images

Figure CN118587748B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of fingerprint recognition technology, and in particular to a method for financial data identity authentication based on fingerprint recognition, which can be used for identity authentication in various scenarios, such as data processing identity authentication in the financial field. Background Art
[0002] With the rapid development of information technology, the financial industry increasingly relies on efficient and secure authentication methods to ensure data confidentiality and integrity. Traditional authentication methods such as usernames and passwords are at risk of being stolen and cracked, so more secure and reliable authentication methods are needed. Fingerprint-based authentication technology has attracted much attention due to its uniqueness and unforgeability. Fingerprint recognition technology analyzes the biometric features of individual fingerprints and converts them into digital information for identification, with high accuracy and reliability. Before performing fingerprint-based authentication, fingerprint image refinement is a necessary operation, the purpose of which is to refine a fingerprint image with a large width into a fingerprint skeleton that only retains a certain center line, so as to improve the speed and accuracy of fingerprint recognition.
[0003] In the existing technology, the OPTA algorithm is a common serial fingerprint thinning algorithm, and its main logic is to use a certain number of templates to delete and retain pixels in the fingerprint binary image, and finally achieve the effect of fingerprint image thinning. However, the setting of the OPTA algorithm template is not appropriate, and the refined fingerprint image will have burrs and other problems, which has a great impact on the feature extraction during fingerprint recognition and the accuracy and recognition speed of subsequent fingerprint recognition, and thus cannot meet the high efficiency of financial data identity authentication based on fingerprint recognition. Summary of the invention
[0004] The present invention provides a financial data identity authentication method based on fingerprint recognition to solve the problem of inaccurate extraction of refined fingerprint features and poor fingerprint recognition accuracy caused by fingerprint refinement burrs in the existing financial data fingerprint identity authentication process. The technical solution adopted is as follows:
[0005] The present invention proposes a financial data identity authentication method based on fingerprint recognition, which includes the following steps:
[0006] Obtain the fingerprint grayscale image and its single pixel width refined fingerprint binary image;
[0007] According to the coordinate distribution of each refined fingerprint pixel in the single pixel width refined fingerprint binary image within the preset neighborhood range, the binary directionality of each refined fingerprint pixel is obtained; according to the similarities and differences between the binary directionality of the refined fingerprint pixel and the directionality of the pixel within the preset domain range, the initial burr dispute parameters of the refined fingerprint pixel are obtained;
[0008] Obtaining the grayscale directionality of the refined fingerprint pixel point according to the minimum grayscale gradient of the pixel point corresponding to the refined fingerprint pixel point in the fingerprint grayscale image and the pixel points within the preset neighborhood range;
[0009] According to the grayscale directionality of the refined fingerprint pixel point and the binary directionality of the refined fingerprint pixel point, a direction difference coefficient is obtained; according to the direction difference coefficient, the initial burr dispute parameter of the refined fingerprint pixel point is corrected to obtain the final burr dispute parameter;
[0010] The single-pixel-width refined fingerprint binary image is processed according to the final burr dispute parameter to obtain a burr-free single-pixel-width refined fingerprint binary image; based on the burr-free single-pixel-width refined fingerprint binary image, combined with the fingerprint data of the financial data authentication fingerprint comparison library, the fingerprint identity authentication result of the person to be authenticated corresponding to the burr-free single-pixel-width refined fingerprint binary image is obtained.
[0011] Optionally, the method of obtaining the fingerprint grayscale image and refining the fingerprint binary image by using the single pixel width thereof includes:
[0012] The original fingerprint data is grayed to obtain a fingerprint grayscale image; the fingerprint grayscale image is processed by Gaussian filtering to obtain a fingerprint filtered grayscale image;
[0013] Perform threshold segmentation on the fingerprint filter grayscale image according to a threshold segmentation algorithm such as the Otsu threshold method to obtain a fingerprint processing binary image of the person to be authenticated;
[0014] According to the OPTA thinning algorithm, the fingerprint binary image of the authenticator is processed and thinned to obtain a single-pixel width thinned fingerprint binary image.
[0015] Optionally, the obtaining of the binary directionality of each refined fingerprint pixel includes the following specific methods:
[0016] For each refined fingerprint pixel point, the coordinates of the refined fingerprint pixel point are subtracted from other refined fingerprint pixels in a preset neighborhood range, and each obtained vector is used as each binary directionality of the corresponding refined fingerprint pixel point.
[0017] Optionally, the obtaining of the initial burr dispute parameters for refining the fingerprint pixel points includes the following specific methods:
[0018] Refine any one of the fingerprint pixels G in the single-pixel-width fingerprint binary image a Mark all the pixels in the row and column of , recorded as pixel point G a The row or column where the
[0019] G a The directionality of all pixels in the row where the pixel is located is fitted to obtain Ga Pixel row fitting direction;
[0020] G a The directionality of all pixels in the column where the pixel is located is fitted to obtain G a Pixel point column fitting direction;
[0021] G a Each directionality of the pixel is related to G a Perform a modulo operation in the row fitting direction or column fitting direction of the pixel point, and select the minimum value of all modulo operations to obtain the refined fingerprint pixel point G a Initial burr dispute parameters.
[0022] Optionally, the grayscale directionality of the refined fingerprint pixel point is obtained, including the following specific methods:
[0023] Take any refined fingerprint pixel G in the grayscale image of the refined fingerprint pixel a As the goal, get G a Grayscale information corresponding to all pixels within a preset neighborhood;
[0024] Will belong to G a The pixel corresponding to the smallest grayscale gradient value among all pixels in the preset neighborhood is G a Direction angle pixel point;
[0025] According to G a The pixel position coordinates and G a The direction angle pixel coordinate information generates a vector and obtains G a Grayscale directionality.
[0026] Optionally, the obtaining of the directional difference coefficient includes a specific method as follows:
[0027] According to G a The inverse cosine similarity is calculated by combining the binary directionality and the grayscale directionality, and the inverse cosine similarity obtained is the refined fingerprint pixel point G a The directional difference coefficient.
[0028] Optionally, the obtaining of the final burr dispute parameter includes the following specific methods:
[0029] According to the grayscale image of the refined fingerprint pixel, obtain any refined fingerprint pixel G a The grayscale directionality inaccuracy coefficient;
[0030] G a The ratio of the directional difference coefficient to the grayscale directional inaccuracy coefficient is used as G a Adjustment parameters of the initial burr dispute parameters;
[0031] G a The initial burr dispute parameter is taken as the main body, and the adjustment parameter of the initial burr dispute parameter is taken as the weighted sum. a The initial burr dispute parameters are adjusted to obtain G a The final burr dispute parameter.
[0032] Optionally, the obtaining of the burr-free single pixel width thinned fingerprint binary image includes the following specific methods:
[0033] According to the single pixel width refinement, the final burr dispute parameter corresponding to each pixel in the fingerprint binary image is refined, and the threshold screening method is used to use the final burr dispute parameter as screening data for each pixel in the single pixel width refinement fingerprint binary image;
[0034] Select a suitable threshold to screen each pixel in all the single-pixel width thinning fingerprint binary images, and select the burr pixel in the single-pixel width thinning fingerprint binary image that is greater than the threshold;
[0035] The grayscale value of the identified burr pixel is changed to 255 to obtain a burr-free single pixel width thinned fingerprint binary image.
[0036] Optionally, the acquisition of any refined fingerprint pixel point G a The grayscale directional inaccuracy coefficient includes the following specific methods:
[0037]
[0038] gap a Represents G a The grayscale directionality inaccuracy coefficient;
[0039] Represents pixel G a The minimum value of the directional gradient within the neighborhood direction;
[0040] It represents the mean of the minimum directional gradients of all pixels with grayscale value G in the neighborhood in the refined fingerprint pixel grayscale image.
[0041] Optionally, the obtaining of the fingerprint identity authentication result of the person to be authenticated corresponding to the burr-free single-pixel width thinned fingerprint binary image includes the following specific methods:
[0042] The burr-free single pixel width thinned fingerprint binary image of the person to be authenticated is used as input data and input into the financial data authentication fingerprint comparison library;
[0043] The fingerprint binary image of the burr-free single pixel width of the person to be authenticated is refined using the fingerprint singular point extraction algorithm to extract features, and compared with all the saved feature points in the financial data authentication fingerprint comparison library;
[0044] The comparison result is output to obtain the fingerprint identity authentication result of the person to be authenticated.
[0045] The beneficial effects of the technical solution of the present invention are as follows: based on the OPTA algorithm, the present invention obtains a refined fingerprint binary image of a single pixel width for a fingerprint grayscale image, and quantifies an initial burr dispute parameter by analyzing the distribution difference of the directionality of the pixel points in the refined fingerprint binary image. The initial burr dispute parameter preliminarily reflects the burr possibility of the pixel points from the refined fingerprint binary image; and further quantifies the grayscale directionality of the refined fingerprint pixel points by analyzing the gradient change of the pixel points in the fingerprint grayscale image, so as to avoid the initial burr dispute parameter from causing an error in the extraction of fingerprint ridges due to the fact that the refined fingerprint binary image contains less physical fingerprint information, thereby affecting fingerprint recognition. Therefore, the directional difference coefficient is quantified based on the grayscale directionality and the binary directionality, and the final burr dispute parameter is corrected to obtain a burr-free single pixel width refined fingerprint binary image and perform fingerprint identity authentication, thereby eliminating the refined fingerprint burr points and ensuring the accuracy and efficiency of financial data identity authentication. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0047] Figure 1 A flowchart of the steps of the financial data identity authentication method based on fingerprint recognition provided by the present invention. DETAILED DESCRIPTION
[0048] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation method, structure, features and effects of the financial data identity authentication method based on fingerprint recognition proposed by the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.
[0049] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
[0050] The general fingerprint recognition method for financial data identity authentication is to input a fingerprint to be identified into an image binarization processing module to obtain a binary fingerprint image that only retains grayscale values 0 and 255, and then process the binarized fingerprint image using the OPTA algorithm to obtain a fingerprint image with a single pixel point of width one, and then input the fingerprint image with a single pixel point of width one into the financial data authentication fingerprint comparison library for feature comparison to obtain the recognition result of the fingerprint to be identified.
[0051] However, when the fingerprint is processed by the OPTA algorithm, when the template is deleted or the template is retained, although the fingerprint to be processed can be thinned into a binary fingerprint image with a pixel width of 1, because the fingerprint itself has a variety of shape features (ridges, valleys, bifurcation points, endpoints, etc.), when the OPTA algorithm is thinning it, some obvious features (such as ridges showing a relatively regular shape) can be effectively thinned, which is specifically reflected in that after the thinning is complete, the corresponding part of the features is a binary image with a single pixel width that is obvious and has no burrs, while some other features cannot be effectively identified (such as the mixed situation of bifurcation points and endpoints), which is specifically reflected in that the corresponding part of the features is a binary image with a single pixel width with a certain degree of burrs, resulting in misjudgment or wrong judgment when comparing features in the subsequent thinned financial data authentication fingerprint comparison library. Therefore, in order to accurately and comprehensively identify fingerprints, it is necessary to process the burrs in the fingerprint thinning process of the OPTA algorithm to obtain a binary image with a single pixel width without burrs.
[0052] The specific scheme of the financial data identity authentication method based on fingerprint recognition provided by the present invention is described in detail below with reference to the accompanying drawings.
[0053] See also Figure 1 , which shows a flowchart of a method for financial data identity authentication based on fingerprint recognition provided by an embodiment of the present invention, the method comprising the following steps:
[0054] Step S001: Obtain a fingerprint grayscale image and a single pixel width refined fingerprint binary image.
[0055] The initial fingerprint image of the person to be authenticated is collected, and the initial fingerprint image of the person to be authenticated is processed using image processing technology to obtain a fingerprint processing binary image of the person to be authenticated; then, based on the fingerprint processing binary image of the person to be authenticated, a single-pixel point width thinning fingerprint binary image is obtained using the OPTA thinning algorithm.
[0056] It should be noted that, because the existing fingerprint thinning algorithms are all based on the binary image of the fingerprint, and the fingerprint image of the person to be authenticated collected by the fingerprint recognition device is a multi-channel color image, it is impossible to perform direct thinning operations, so it is necessary to perform image processing on the collected fingerprint to obtain the fingerprint processed binary image of the person to be authenticated.
[0057] As a preferred example, the process of obtaining the fingerprint processing binary image to be authenticated is as follows:
[0058] First, the corresponding finger of the person to be identified is placed on the optical lens, and then the corresponding finger of the person to be identified is illuminated by the built-in illumination light source of the optical identification device. Then, according to the reflection of the light on the uneven lines of the fingerprint on the surface of the finger, the multi-grayscale value fingerprint image with ridges and valleys of the person to be identified is obtained as the original data by using the charge coupled device CCD.
[0059] Secondly, the original data is grayed to obtain the next level processed image;
[0060] Then, the second-level processed image is processed by Gaussian filtering to obtain a third-level image;
[0061] It should be noted that the significance and purpose of Gaussian filtering for the next level of image processing is to blur the original data, so that unnecessary defects in the original data can be completed and unnecessary endpoints can be removed, so as to improve the overall quality of the original data.
[0062] Finally, a threshold segmentation algorithm such as the Otsu threshold method is used to perform threshold segmentation on the second-level image to obtain a fingerprint processed binary image of the person to be authenticated.
[0063] It should be noted that when the fingerprint processing binary image of the person to be authenticated obtained by the optimization example is used for subsequent identity authentication, the overall data quality is high, which makes the subsequent refinement burr processing process less computationally intensive and the burr processing more accurate.
[0064] As an optional example, the process of obtaining the binary image of the fingerprint processing of the person to be authenticated is as follows:
[0065] Firstly, a threshold segmentation algorithm such as Otsu threshold method is used to perform threshold segmentation on the secondary processed image to obtain a tertiary processed image;
[0066] Finally, the three processed images are processed using a morphological algorithm such as dilation and erosion to obtain a processed binary image of the fingerprint of the person to be authenticated.
[0067] As another optional example, the process of obtaining the binary image of the fingerprint of the person to be authenticated is as follows:
[0068] First, a threshold segmentation algorithm such as the Otsu threshold method is used to perform threshold segmentation on the original image to obtain a fingerprint processed binary image of the person to be authenticated.
[0069] It should be noted that the optional example processes the fingerprint of the person to be authenticated to obtain the fingerprint processing binary image of the person to be authenticated faster overall, and is more suitable for scenarios where authentication is frequent or the hardware foundation of the device is relatively weak.
[0070] It needs to be further explained that when the person to be authenticated performs fingerprint recognition-based identity authentication in the financial data authentication fingerprint comparison library, it is necessary to use a burr-free single-pixel point width refined fingerprint binary image to perform feature comparison with the fingerprints in the financial data authentication fingerprint comparison library to obtain the identity authentication result. Therefore, it is first necessary to obtain a single-pixel point width refined fingerprint binary image of the person to be authenticated to facilitate subsequent burr removal processing.
[0071] As an example, the method for obtaining the single-pixel width thinned fingerprint binary image of the person to be authenticated is as follows:
[0072] The OPTA thinning algorithm is used to perform sliding window processing on the fingerprint binary image of the authenticated person based on retaining templates and deleting templates to obtain a single-pixel width thinned fingerprint binary image of the authenticated person.
[0073] Step S002: According to the coordinate distribution of each refined fingerprint pixel in the single-pixel width refined fingerprint binary image within a preset neighborhood range, the binary directionality of each refined fingerprint pixel is obtained; according to the similarities and differences between the binary directionality of the refined fingerprint pixel and the directionality of the pixel within the preset domain range, the initial burr dispute parameters of the refined fingerprint pixel are obtained.
[0074] In the single-pixel width thinned fingerprint binary image thinned by the OPTA algorithm, there are a certain number of burr points, and the presence of burr points will affect the acquisition of feature points when performing fingerprint feature matching, thereby making the fingerprint recognition result inaccurate. In actual physical reality, fingerprints have certain directional characteristics. Even the single-pixel width thinned fingerprint binary image after being thinned by the OPTA algorithm has distinct directional characteristics. Generally speaking, the burrs and feature points (fingerprint endpoints, bifurcation points) and other pixel points in the single-pixel width thinned fingerprint binary image are not satisfied with the directional characteristics of the fingerprint and are highly controversial. In this embodiment, the single-pixel width thinned fingerprint binary image is obtained by sliding window pixel neighborhood feature statistics to obtain the directionality of each pixel in the single-pixel width thinned fingerprint binary image, and then analyzes the similarities and differences between the directionality of each pixel and the directionality of the pixels within a certain range, thereby indicating the initial burr dispute parameters of each pixel in the single-pixel width thinned fingerprint binary image.
[0075] It should be noted that the initial burr dispute parameter of each pixel of the single-pixel width thinned fingerprint binary image does not represent the final calculation amount of whether each pixel of the single-pixel width thinned fingerprint binary image is a burr pixel, because the initial burr dispute parameter of each pixel of the single-pixel width thinned fingerprint binary image is obtained through the single-pixel width thinned fingerprint binary image, and the amount of information contained in the binary image is essentially small, so it is impossible to effectively judge whether each pixel in the single-pixel width thinned fingerprint binary image is truly a burr pixel. And in this embodiment, the main purpose of this parameter is to perform an initial screening for each pixel in the single-pixel width thinned fingerprint binary image, and to roughly judge the concept of each pixel on the single-pixel width thinned fingerprint binary image as a burr pixel by using the binary image containing less information.
[0076] As a preferred example, the method for obtaining the binary directionality of each refined fingerprint pixel is as follows:
[0077] The ath pixel in the single-pixel-width thinned fingerprint binary image is taken as the starting pixel, denoted as G a ;
[0078] G a As the positioning pixel, a 3*3 window is applied, where G a The pixel point is the center pixel point of the window;
[0079] G a Subtract the coordinates of each refined fingerprint pixel in the window, and each vector obtained is G a Each directionality of a pixel.
[0080] As another preferred example, the method for obtaining the initial burr dispute parameters of each refined fingerprint pixel point is as follows:
[0081] Refine any pixel point G in the fingerprint binary image with a single pixel width a Mark all the pixels in the row and column of , recorded as pixel point G a The row or column where the
[0082] G a The directionality of all pixels in the row where the pixel is located is fitted to obtain G a Pixel row fitting direction;
[0083] G a The directionality of all pixels in the column where the pixel is located is fitted to obtain G a Pixel point column fitting direction;
[0084] G a Each directionality of the pixel is related to Ga Perform a modulo operation in the row fitting direction or column fitting direction of the pixel point, and select the minimum value of all modulo operations to obtain the refined fingerprint pixel point G a Initial burr dispute parameters.
[0085] In the above, it is considered that in the single-pixel width thinned fingerprint binary image, due to the distribution of the fingerprint ridges, more thinned fingerprint pixels are in the same direction in the same row or column, so the directional fitting of the pixels in the same row or column is consistent with the directional distribution of most pixels in the current row or column. a Calculate the fitting direction of all fingerprint pixels in the same row or column G a The initial burr dispute parameters are selected, and the minimum value is selected as G a The reason for the controversial parameter of the initial burr is that G a The closer it is to the row fitting direction or column fitting direction, the more G a The closer the direction of G is to the row fitting direction or column fitting direction, the greater the avoidance of a Misjudgment of initial burr parameters due to selecting the wrong fitting direction.
[0086] Step S003: Obtain the grayscale directionality of the refined fingerprint pixel point according to the grayscale gradient minimum value of the pixel point corresponding to the refined fingerprint pixel point in the fingerprint grayscale image and the pixel points within a preset neighborhood.
[0087] In the above logic, the initial burr dispute parameters of each refined fingerprint pixel point are obtained only by refining the fingerprint binary image with a single pixel width, and the physical information about the fingerprint of the person to be authenticated carried by the binary image is actually relatively small, so it is impossible to accurately determine whether a pixel point in a single pixel width refined fingerprint binary image is a burr pixel point based on the initial burr dispute parameters alone. The single pixel width refined fingerprint binary image is actually obtained by reducing the dimensionality of the information related to the ridges in the grayscale image of the fingerprint of the person to be authenticated, but compared with the single pixel width refined fingerprint binary image, the grayscale image of the fingerprint of the person to be authenticated contains more fingerprint information, such as the direction information of the obvious ridges, so this embodiment performs grayscale directional analysis on some pixels in the grayscale image of the fingerprint of the person to be authenticated based on the grayscale image of the fingerprint of the person to be authenticated.
[0088] It should be noted that, in the implementation process of this embodiment, firstly, the binary directionality of each pixel in the fingerprint binary image with a single pixel width is obtained, and then the grayscale directionality of some pixels in the fingerprint grayscale image of the person to be authenticated is obtained. In essence, it is not a repeated operation, because only the grayscale directionality of the fingerprint grayscale image of the person to be authenticated is used to obtain the grayscale directionality. Without preprocessing the fingerprint grayscale image of the person to be authenticated, we cannot obtain the pixels in the fingerprint grayscale image of the person to be authenticated that are the ridge pixels that we need to represent most of the fingerprint information. Therefore, it is necessary to calculate and process each pixel in the entire fingerprint grayscale image of the person to be authenticated. The overall calculation amount is not only large, but also the overall calculation process is relatively redundant. However, the binary directionality of each pixel in the fingerprint binary image with a single pixel width is first obtained by using the present embodiment; then the fingerprint binary image with a single pixel width is projected onto the fingerprint grayscale image of the person to be authenticated, and the grayscale directionality of some pixels in the projection effect is obtained, which can reduce a lot of calculations.
[0089] As a preferred example, the method for obtaining a grayscale image of refined fingerprint pixels is as follows:
[0090] The position coordinates of each pixel in the single-pixel width thinned fingerprint binary image are counted;
[0091] Project the grayscale image of the fingerprint of the person to be authenticated according to the position coordinates of each pixel point in the binary image of the fingerprint refined according to the statistics of the single pixel width to obtain the grayscale image of the refined fingerprint pixel point;
[0092] The pixel point in the original fingerprint grayscale image of the person to be authenticated after projection and the same coordinate position in the single-pixel-width refined fingerprint binary image is called the refined fingerprint pixel point.
[0093] As another preferred example, the method for obtaining the grayscale directionality of the refined fingerprint pixel point is as follows:
[0094] Take any refined fingerprint pixel G in the grayscale image of the refined fingerprint pixel a As the goal, get G a Grayscale information corresponding to all pixels in the 8-neighborhood;
[0095] Calculate G a With G a The grayscale gradient information corresponding to all pixels within the eight neighborhoods is obtained. a With G a The gradient value of each pixel in the eight-neighborhood;
[0096] Choose to belong to G a The smallest gradient value among the eight gradient values of G a Direction angle;
[0097] G a The pixel position coordinates and G a The direction angle pixel coordinate information generates a vector and obtains G a Grayscale directionality.
[0098] In the above, it is considered that in the fingerprint grayscale image, the fingerprints with ridge features should have grayscale values that are close to each other, so G a The gray gradient is calculated with the pixels in the neighborhood. The smaller the gray gradient is, the smaller the pixel G is. a The smaller the gray value difference between the corresponding neighboring pixel in this direction is, the smaller the gray value difference between the corresponding neighboring pixel in this direction is. a is a ridge pixel, then the neighboring pixel with a smaller grayscale difference value is more likely to be a ridge pixel, and the ridge pixels constitute the fingerprint, so G is selected. a The direction with the smallest gradient with the neighboring pixels in the grayscale image is regarded as the grayscale directionality.
[0099] Step S004: obtaining a direction difference coefficient according to the grayscale directionality of the refined fingerprint pixel and the binary directionality of the refined fingerprint pixel; and correcting the initial burr dispute parameters of the refined fingerprint pixel according to the direction difference coefficient to obtain the final burr dispute parameters.
[0100] In history, the binary directionality of the obtained refined fingerprint pixel point has been explained, and the overall accuracy is not enough. It can only analyze the possibility of each pixel point as a burr pixel point from the perspective of the binary image (i.e., the spatial position distribution angle of the pixel point); and in the process of obtaining the grayscale directionality of the refined fingerprint pixel point, only the grayscale feature connection of the refined fingerprint pixel point is emphasized, and the spatial position feature of the refined fingerprint pixel point is not effectively paid attention to. There are limitations in using any one of these two features to determine whether the pixel point in the single pixel width refined fingerprint binary image is a burr pixel point of the refined fingerprint. Therefore, this embodiment compares the binary directionality corresponding to the fingerprint image of the person to be authenticated with the grayscale directionality to obtain the direction difference coefficient, and then uses the direction difference coefficient to correct the binary feature, i.e., the initial burr dispute parameter, to obtain the final burr dispute parameter. By integrating the common features of the fingerprint image of the person to be authenticated at the binary image angle and the grayscale image angle, the burr fingerprint pixel point on the single pixel width refined fingerprint binary image is further and more accurately judged.
[0101] As a preferred example, the method for obtaining the directional difference coefficient is as follows:
[0102] First, obtain any refined fingerprint pixel G a The directionality of the fingerprint images in two forms is G aThe binary directionality and grayscale directionality of
[0103] Secondly, using G a The inverse cosine similarity is calculated by combining the binary directionality and the grayscale directionality, and the inverse cosine similarity is the refined fingerprint pixel point G a Directional difference coefficient;
[0104] It should be noted that G a The binary directionality is G a Directionality of the initial burr dispute parameters involved in the calculation.
[0105] It should be noted that the directional difference coefficient is calculated by calculating the directional difference value of the refined fingerprint pixel in the grayscale image of the refined fingerprint pixel and the single pixel width refined fingerprint binary image. The smaller the value obtained by the calculation, the smaller the refined fingerprint pixel G. a The more accurate the directionality obtained in the grayscale image of the refined fingerprint pixel points, the more accurate the initial burr dispute parameters obtained based on the grayscale image of the refined fingerprint pixel points. Otherwise, the initial burr dispute parameters are not accurate enough and further judgment and correction are required.
[0106] As a preferred example, the method for obtaining the final burr dispute parameters is as follows:
[0107] Use the grayscale image of the refined fingerprint pixel to obtain any refined fingerprint pixel G a The grayscale directionality inaccuracy coefficient;
[0108] G a The ratio of the directional difference coefficient to the grayscale directional inaccuracy coefficient is taken as G a Adjustment parameters of the initial burr dispute parameters;
[0109] G a The initial burr dispute parameters are taken as the main body, and G is adjusted by using the adjustment parameters as the weighted sum a The initial burr dispute parameters are adjusted to obtain G a The final burr dispute parameter.
[0110] It should be noted that the final burr dispute parameter is a characteristic parameter used to determine whether the refined fingerprint pixel is a burr pixel. Compared with the existing technology, in this embodiment, by a Based on the binary directionality, G aThe initial burr parameter is the preliminary judgment condition of the burr pixel point; then, based on the grayscale direction accuracy, the directional difference parameters of the grayscale direction and the binary direction are used to jointly verify and adjust the initial burr parameter, so that the G can be judged more accurately from the perspectives of the single pixel width refinement of the fingerprint binary image and the refinement of the fingerprint pixel grayscale image. a Whether it is a glitch pixel.
[0111] Step S005: Process the single-pixel width refined fingerprint binary image according to the final burr dispute parameter to obtain a burr-free single-pixel width refined fingerprint binary image; based on the burr-free single-pixel width refined fingerprint binary image, combined with the fingerprint data of the financial data authentication fingerprint comparison library, obtain the fingerprint identity authentication result of the person to be authenticated corresponding to the burr-free single-pixel width refined fingerprint binary image.
[0112] The final burr dispute parameter mentioned above is a judgment parameter for whether each refined fingerprint pixel in the single-pixel width refined fingerprint binary image is a burr pixel. When the fingerprint image of the person to be authenticated is to be recognized, it is necessary to use the burr-free single-pixel width refined fingerprint binary image for feature extraction and feature comparison, and finally output the recognition result. Therefore, the specific method of this embodiment is as follows:
[0113] As an example, the method for obtaining a burr-free single pixel width thinned fingerprint binary image is as follows:
[0114] First, the final burr dispute parameter corresponding to each pixel in the single pixel width thinned fingerprint binary image is obtained;
[0115] Secondly, the threshold screening method is used to refine each pixel in the fingerprint binary image based on the single pixel width according to the final burr dispute parameter as the screening data;
[0116] Then, a suitable threshold is selected to screen each pixel in all the single-pixel width thinning fingerprint binary images, and the burr pixel in the single-pixel width thinning fingerprint binary image with a width greater than the threshold is selected;
[0117] Finally, the grayscale value of the identified burr pixel is changed to 255, and the burr-free single pixel width thinned fingerprint binary image can be obtained;
[0118] As an example, the method for obtaining the fingerprint authentication result of the person to be authenticated is as follows:
[0119] First, a burr-free single-pixel-width thinned fingerprint binary image of the person to be authenticated is selected as input data and input into the financial data authentication fingerprint comparison library;
[0120] Secondly, the fingerprint binary image of the burr-free single pixel width of the person to be authenticated is refined using the fingerprint singular point extraction algorithm to extract features, and compared with all the saved feature points in the financial data authentication fingerprint comparison library;
[0121] Finally, the comparison result is output to obtain the fingerprint identity authentication result of the person to be authenticated.
[0122] Embodiment 2:
[0123] In this embodiment, as a preferred example, the method for obtaining the initial burr dispute parameter of each refined fingerprint pixel point is:
[0124] There are multiple refined fingerprint pixels in the single pixel width refined fingerprint binary image, but not every pixel is the pixel corresponding to the burr feature after fingerprint refinement. Obviously, fingerprints have directional characteristics, so in this embodiment, the directional characteristics of different refined fingerprint pixels are used to quantify each initial burr dispute parameter.
[0125] As an example, the specific formula for the initial burr dispute parameter of the a-th pixel in the single-pixel width thinning fingerprint binary image is:
[0126]
[0127] IP a Represents the initial burr dispute parameter of the a-th pixel in the single-pixel width thinned fingerprint binary image;
[0128] Represents the refined fingerprint pixel G a All N directions, with pixel G a The fitting directionality of all pixels in the same row The minimum value after the modulo operation;
[0129] Represents the refined fingerprint pixel G a All N directions, with pixel G a Fitting directionality of all pixels in the same column The minimum value after the modulo operation;
[0130] represents the limited weight of the initial burr dispute parameter of the a-th pixel;
[0131] δ Ha Represents the refined fingerprint pixel G a The variance of the cosine similarity of the directions of all adjacent pixels in the same row;
[0132] δLa Represents the refined fingerprint pixel G a The variance of the cosine similarity of the directions of all adjacent pixels in the same column;
[0133] It should be noted that the significance of limiting the weights is to further accurately calculate the initial burr parameters through the specific fingerprint distribution characteristics of the rows and columns where the a-th pixel is located. The basis for this is that although the finger fingerprint has a high directional feature, not every row or column of pixels has obvious directionality. The direction of the local fingerprint may change dramatically. At this time, when the directionality of the fitted row or column pixels is used to refer to the directionality of the a-th pixel, the overall reference effect is poor. Therefore, it is necessary to use limited weights for effective adjustment, so that when calculating the initial burr dispute parameters, the calculation deviation caused by the unclear directionality of the local fingerprint is reduced.
[0134] Embodiment three:
[0135] In this embodiment, the method for obtaining the direction difference coefficient according to the grayscale directionality of the refined fingerprint pixel point and the binary directionality of the refined fingerprint pixel point is:
[0136] As an example, the specific formula of the directional difference coefficient is:
[0137]
[0138] DC a It represents the directional difference coefficient between the binary directionality of the a-th pixel in the single-pixel-width thin fingerprint binary image and the grayscale directionality of the a-th pixel in the grayscale image of the thin fingerprint pixel. COS() is the cosine similarity function.
[0139] represents the representative binary directionality of the a-th pixel in the single-pixel-width thinned fingerprint binary image, Represents the pixel G in the refined fingerprint grayscale image a Grayscale directionality.
[0140] Embodiment 4:
[0141] In this embodiment, the method of correcting the initial burr dispute parameters of the refined fingerprint pixel points according to the directional difference coefficient to obtain the final burr dispute parameters is:
[0142] As an example, the specific formula of the final burr dispute parameter is:
[0143]
[0144] FP aRepresents the final burr dispute parameter of the a-th pixel in the single-pixel-width thinned fingerprint binary image;
[0145] gap a Represents the refined fingerprint pixel G a In the refined fingerprint pixel grayscale image, the grayscale directionality inaccuracy coefficient;
[0146] It should be noted that the final burr dispute parameter is calculated using G a The initial burr dispute coefficient is taken as the main body, and then the directional difference coefficient is adjusted by the inaccurate coefficient of grayscale directionality. The specific logical process is: a The more consistent the directionality is between the refined fingerprint pixel grayscale image and the single pixel width refined fingerprint binary image, that is, DC a The closer it is to 0, the smaller the G obtained by refining the fingerprint binary image with a single pixel width. a The more accurate the initial burr dispute coefficient is, the more accurate the initial burr dispute coefficient is. a Is it a single pixel width of the fingerprint to be authenticated? The higher the accuracy of the burr pixel point of the fingerprint binary image, the higher the accuracy of the burr pixel point of the fingerprint binary image. a When adjusting the features of the fingerprint pixel grayscale image, the adjustment should be smaller; otherwise, it means that the initial burr dispute coefficient is used for G a Whether it is a single-pixel width thinning of the fingerprint to be authenticated, the lower the accuracy of the burr pixel points of the fingerprint binary image, the more it needs to be adjusted significantly under the premise of accurate grayscale directionality.
[0147] As another example, the specific formula of the inaccuracy coefficient of the grayscale directionality of the refined fingerprint pixel point is:
[0148]
[0149] gap a Represents G a The grayscale directionality inaccuracy coefficient;
[0150] Represents pixel G a The minimum value of the directional gradient within the neighborhood direction;
[0151] In the refined fingerprint pixel grayscale image, all pixels with grayscale values of G (i.e. a The mean of the minimum directional gradients of pixels with the same grayscale value in the neighborhood;
[0152] It should be noted that in the refined fingerprint pixel grayscale image, the pixels with the same grayscale value of G should have a relatively consistent minimum gradient, that is, in the neighborhood of all pixels with the grayscale value of G, the grayscale value changes relatively slowly, and the specific change is the fingerprint ridge direction of the person to be authenticated, so G a It follows this change, and then there is G obtained by refining the grayscale image of fingerprint pixels. a The grayscale directionality of G is more accurate, so a The directional difference coefficient is more objective, so when adjusting the final burr dispute parameters, more attention should be paid to G. a The directional difference coefficient of G a The directional difference coefficient is not very objective. When adjusting the final burr dispute parameters, G should be weakened. a Adjustment of the directional difference coefficient.
[0153] This embodiment is now completed. Compared with the financial data identity authentication method based on fingerprint recognition, this embodiment avoids the interference of the burr phenomenon in the binary image after the fingerprint is refined on the fingerprint recognition of the person to be authenticated to a certain extent.
[0154] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A financial data identity authentication method based on fingerprint recognition, characterized in that: The method comprises the following steps: Obtain the fingerprint grayscale image and its single pixel width refined fingerprint binary image; According to the coordinate distribution of each refined fingerprint pixel in the single pixel width refined fingerprint binary image within the preset neighborhood range, the binary directionality of each refined fingerprint pixel is obtained; according to the similarities and differences between the binary directionality of the refined fingerprint pixel and the directionality of the pixel within the preset domain range, the initial burr dispute parameters of the refined fingerprint pixel are obtained; Obtaining the grayscale directionality of the refined fingerprint pixel point according to the minimum grayscale gradient of the pixel point corresponding to the refined fingerprint pixel point in the fingerprint grayscale image and the pixel points within the preset neighborhood range; According to the grayscale directionality of the refined fingerprint pixel point and the binary directionality of the refined fingerprint pixel point, a direction difference coefficient is obtained; according to the direction difference coefficient, the initial burr dispute parameter of the refined fingerprint pixel point is corrected to obtain the final burr dispute parameter; The single pixel width thinned fingerprint binary image is processed according to the final burr dispute parameter to obtain a burr-free single pixel width thinned fingerprint binary image; based on the burr-free single pixel width thinned fingerprint binary image, combined with the fingerprint data of the financial data authentication fingerprint comparison library, the fingerprint identity authentication result of the person to be authenticated corresponding to the burr-free single pixel width thinned fingerprint binary image is obtained; The specific method of obtaining the direction difference coefficient includes: according to The inverse cosine similarity is calculated by combining the binary directionality and the grayscale directionality, and the inverse cosine similarity obtained is the refined fingerprint pixel point Directional difference coefficient; The specific method of obtaining the final burr dispute parameter includes: According to the grayscale image of the refined fingerprint pixel, obtain any refined fingerprint pixel The grayscale directionality inaccuracy coefficient; Will The ratio of the directional difference coefficient to the grayscale directional inaccuracy coefficient is given as Adjustment parameters of the initial burr dispute parameters; by The initial burr dispute parameter is taken as the main body, and the adjustment parameter of the initial burr dispute parameter is taken as the weighted sum. The initial burr dispute parameters are adjusted to obtain The final burr dispute parameter.
2. The method for financial data identity authentication based on fingerprint recognition according to claim 1, characterized in that: The specific method of obtaining the fingerprint grayscale image and refining the fingerprint binary image with a single pixel width includes: The original fingerprint data is grayed to obtain a fingerprint grayscale image; the fingerprint grayscale image is processed by Gaussian filtering to obtain a fingerprint filtered grayscale image; Perform threshold segmentation on the fingerprint filter grayscale image according to the threshold segmentation algorithm to obtain the fingerprint processing binary image of the person to be authenticated; According to the OPTA thinning algorithm, the fingerprint binary image of the authenticator is processed and thinned to obtain a single-pixel width thinned fingerprint binary image.
3. The method for financial data identity authentication based on fingerprint recognition according to claim 1, characterized in that: The specific method of obtaining the binary directionality of each refined fingerprint pixel point includes: For each refined fingerprint pixel point, the coordinates of the refined fingerprint pixel point are subtracted from other refined fingerprint pixels in a preset neighborhood range, and each obtained vector is used as each binary directionality of the corresponding refined fingerprint pixel point.
4. The method for financial data identity authentication based on fingerprint recognition according to claim 1, characterized in that: The specific method of obtaining the initial burr dispute parameters of the refined fingerprint pixel points includes: Refine any pixel point of the fingerprint in the single-pixel-width fingerprint binary image Mark all the pixels in the row and column of The row or column where the Will The directionality of all pixels in the row where the pixel is located is fitted to obtain Pixel row fitting direction; Will The directionality of all pixels in the column where the pixel is located is fitted to obtain Pixel point column fitting direction; Will Each directionality of a pixel is Perform a modulo operation in the row fitting direction or column fitting direction of the pixel point, and select the minimum value of all modulo operations to obtain the refined fingerprint pixel point Initial burr dispute parameters.
5. The method for financial data identity authentication based on fingerprint recognition according to claim 1, characterized in that: The specific method of obtaining the grayscale directionality of the refined fingerprint pixel point includes: Any refined fingerprint pixel in the grayscale image of the refined fingerprint pixel As a goal, obtain Grayscale information corresponding to all pixels within a preset neighborhood; Will belong to The pixel corresponding to the smallest grayscale gradient value among all pixels in the preset neighborhood is used as Direction angle pixel point; according to The pixel position coordinates and The direction angle pixel coordinate information generates a vector and obtains Grayscale directionality.
6. The method for financial data identity authentication based on fingerprint recognition according to claim 1, characterized in that: The specific method of obtaining a burr-free single pixel width thinned fingerprint binary image includes: According to the single pixel width refinement, the final burr dispute parameter corresponding to each pixel in the fingerprint binary image is refined, and the threshold screening method is used to use the final burr dispute parameter as screening data for each pixel in the single pixel width refinement fingerprint binary image; Select a suitable threshold to screen each pixel in all the single-pixel width thinning fingerprint binary images, and select the burr pixel in the single-pixel width thinning fingerprint binary image that is greater than the threshold; The grayscale value of the identified burr pixel is changed to 255 to obtain a burr-free single pixel width thinned fingerprint binary image.
7. The method for financial data identity authentication based on fingerprint recognition according to claim 1, characterized in that: The method of obtaining any refined fingerprint pixel point The grayscale directional inaccuracy coefficient includes the following specific methods: express The grayscale directionality inaccuracy coefficient; Represents pixel The minimum value of the directional gradient within the neighborhood direction; Indicates that in the refined fingerprint pixel grayscale image, all grayscale values are The average of the minimum directional gradients of the pixel points in the neighborhood.
8. The method for financial data identity authentication based on fingerprint recognition according to claim 1, characterized in that: The specific method of obtaining the fingerprint identity authentication result of the person to be authenticated corresponding to the burr-free single pixel width thinned fingerprint binary image includes: The burr-free single pixel width thinned fingerprint binary image of the person to be authenticated is used as input data and input into the financial data authentication fingerprint comparison library; The fingerprint binary image of the burr-free single pixel width of the person to be authenticated is refined using the fingerprint singular point extraction algorithm to extract features, and compared with all the saved feature points in the financial data authentication fingerprint comparison library; The comparison result is output to obtain the fingerprint identity authentication result of the person to be authenticated.
9. The method for financial data identity authentication based on fingerprint recognition according to claim 2, characterized in that: The step of performing threshold segmentation on the fingerprint filtered grayscale image according to the threshold segmentation algorithm comprises: The fingerprint filtered grayscale image is threshold segmented according to the Otsu threshold method.
Citation Information
Patent Citations
An electric vehicle fingerprint identification intelligent anti-theft system
CN108974192A
Fingerprint image enhancement method based on knowledge
CN1421815A
