Iris information amount and complex number domain feature low-computing-power scene identity verification method

Abstract

According to the iris information amount and complex field feature low-computing-power scene identity verification method, after feature data of template iris convolution data of a template tester after convolution processing is carried out, the optimal partitioning dimension is obtained through an analysis result of each sub-block based on the information amount of the iris so as to partition an image, and then the feature data of the template iris convolution data of the template tester are obtained; the method comprises the following steps: determining whether a stable relationship exists between sub-blocks according to a preset threshold value, further determining whether the sub-blocks with the stable relationship are a constant relationship, performing two-stage verification on test iris feature data of a tester based on an established constant relationship set, and finally determining whether the tester belongs to a template tester. According to the method, a segmentation scheme which can best reflect iris texture features is accurately selected through a texture region guide segmentation strategy based on the information amount, traditional real number field feature expression is expanded to a complex number space, complex changes of iris textures can be more comprehensively described, the problem of uncertainty of iris image acquisition is effectively solved, and the iris image acquisition accuracy is improved. And the practicability and adaptability of the system are enhanced.

Description

Technical Field

[0001] This invention relates to the field of identity verification technology, and in particular to an identity verification method for low-computing-power scenarios that combines iris information content with complex field features. Background Technology

[0002] Currently, with the rapid development of the social information industry, the demand for identity verification has increased dramatically. In addition to traditional username and password login verification systems, biometric technologies such as facial recognition, fingerprint recognition, and iris recognition are being widely used, for example, in bank and company attendance systems. Iris recognition, due to its physiological characteristics being difficult to replicate, is more easily applied in highly confidential and high-security locations such as military bases. Current iris authentication methods require test samples and training samples to undergo a unified data processing flow to ensure the consistency of feature data during the recognition process, which is a prerequisite for accurate identification.

[0003] In practical applications, uncertainties exist in the iris image acquisition process, such as differences in acquisition equipment, changes in the external environment, and changes in the behavior of the subject being acquired. These factors make it difficult to effectively express iris feature information, which seriously affects the accuracy of recognition. In addition, traditional iris recognition technology relies on high-performance computing resources. In low-computing-power equipment environments with limited computing and storage resources, traditional methods are difficult to apply widely.

[0004] Therefore, how to accurately identify uncertain iris images, perform more complete feature representation, and complete the matching process under low computing power has become a key problem that urgently needs to be solved in the field of iris recognition. Summary of the Invention

[0005] This invention discloses an identity verification method for low-computing-power scenarios based on iris information content and complex field features, in order to overcome the above-mentioned technical problems.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows:

[0007] A low-computing-power authentication method that combines iris information content and complex field features includes the following steps:

[0008] S1: Obtain X template iris grayscale images of multiple template testers as training data, and normalize the template iris grayscale images to obtain template iris normalized enhanced images; where X represents the total number of template iris grayscale images used as training data.

[0009] S2: Based on the template iris normalization enhancement image, obtain template iris convolution data using gradient Laplacian convolution kernels;

[0010] S3: Randomly set the I-th group block dimension, obtain the feature data of the template iris convolution data in the sub-blocks under the i-th, i=1,…,I-th group block dimension, so as to obtain the g-th... i -1 sub-block, g-th i Sub-blocks and the g-th i The relationships between the feature data of the template iris convolution data in +1 sub-blocks; where i represents the index number of the group in the block dimension; I represents the total number of groups in the block dimension; g i This represents the index number of the sub-block under the i-th group's block dimension;

[0011] S4: Based on the block dimension of the i-th, i=1,…,I group, the g-th i -1 sub-block, g-th i Sub-blocks and the g-th i The relationship between the feature data of the template iris convolutional data in the +1 sub-blocks is obtained to obtain the analysis results of the information content of the sub-blocks based on the iris under the block dimension of the i, i = 1, ..., I group, so as to obtain the optimal block dimension of the template iris convolutional data, and then determine the feature data of the template iris convolutional data in the sub-blocks under the optimal block dimension.

[0012] S5: Based on the feature data of the template iris convolution data within the sub-blocks under the optimal block dimension, determine whether the g-th sub-block and the g'-th sub-block under the optimal block dimension are stable; where g and g' are the index numbers of the sub-blocks under the optimal block dimension.

[0013] S6: When the g-th sub-block and the g'-th sub-block under the optimal partitioning dimension are in a stable relationship, determine whether the g-th sub-block and the g'-th sub-block under the optimal partitioning dimension are in a constant relationship in order to establish a constant relationship set;

[0014] S7: When the g-th sub-block and the g'-th sub-block have a constant relationship under the optimal block segmentation dimension, obtain the complex features of the sub-blocks that have a constant relationship, and then obtain the complex features of the template testers;

[0015] S8: Obtain the grayscale image of the test iris of the tester to obtain the normalized enhanced image of the test iris, and then obtain the convolutional data of the test iris;

[0016] S9: Based on the optimal segmentation dimension, obtain the test iris feature data of the g-th sub-block; to determine whether the test iris feature data of the g-th sub-block may belong to the template tester;

[0017] S10: When the test iris feature data of the g-th sub-block may belong to the template tester, obtain the real part representation of the complex feature of the test iris feature data of the g-th sub-block to obtain the complex feature of the tester;

[0018] S11: Based on the multiple characteristics of template testers and multiple characteristics of testers, determine whether a tester belongs to the template tester, and realize identity verification based on iris information.

[0019] Beneficial Effects: The present invention provides a low-computing-power scenario identity verification method based on iris information content and complex domain features. After analyzing the feature data of the template iris convolution data of the template tester after convolution processing, the method obtains the optimal segmentation dimension to segment the image based on the information content of each sub-block, determines whether there is a stable relationship between the sub-blocks, and then determines whether the sub-blocks with stable relationships are constant. Based on the established set of constant relationships, the method performs two-stage verification on the tester's test iris feature data, and finally determines whether the tester belongs to the template tester.

[0020] This invention employs an information-based texture region-guided segmentation strategy to accurately select the segmentation scheme that best reflects the iris texture features, effectively overcoming the limitations of traditional methods that directly describe texture changes to obtain iris information. This improves the scientific rigor and accuracy of iris texture information analysis, reduces data redundancy, and lowers the demand for computing resources.

[0021] This invention also extends the traditional real number field feature representation to the complex number space, which can more delicately and comprehensively depict the complex changes in iris texture, break through the inherent limitations of real number field representation, and improve the richness and diversity of features.

[0022] By organically combining texture-based variation patterns with complex spatial feature representation, this method effectively addresses the uncertainties inherent in iris image acquisition. It ensures a constant relative relationship between dimensions even when iris image data fluctuates, scientifically guaranteeing the stability and reliability of feature matching. Even when processing complex and variable iris images, it maintains high recognition performance, enhancing the system's practicality and adaptability, and providing a reliable solution for iris recognition in resource-constrained environments. This effectively addresses the uncertainties in iris image acquisition, enhancing the system's practicality and adaptability. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart of the identity verification method of the present invention;

[0025] Figure 2 This is a schematic diagram of the overall operation process of an embodiment of the present invention;

[0026] Figure 3 This is a schematic diagram illustrating the experimental results of the verification method of this invention and the prior art method based on the JLU iris database.

[0027] Figure 4 This diagram illustrates the experimental results of the verification method of this invention and existing technical methods based on the CASIA iris database. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. 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.

[0029] This embodiment introduces a low-computing-power authentication method based on iris information content and complex field features, including the following steps: Figure 1 , Figure 2 As shown:

[0030] S1: Obtain X template iris grayscale images of multiple template testers as training data, and normalize the template iris grayscale images to obtain template iris normalized enhanced images; where X represents the total number of template iris grayscale images used as training data.

[0031] Specifically, this embodiment uses a commercially available iris acquisition device to acquire 1,000 template iris grayscale images of multiple template testers; and uses the Daugman rubber band method and equalization histogram to convert all template iris grayscale images into 256×32 dimension template iris normalization enhancement images.

[0032] S2: Based on the template iris normalization enhancement image, obtain template iris convolution data using gradient Laplacian convolution kernels;

[0033] Specifically, all template iris normalization enhancement images of the template testers are convolved using gradient Laplacian convolution kernels. The grayscale value of each template iris normalization enhancement image after convolution is read to obtain 256×32 dimension template iris convolution data.

[0034] S3: Randomly set I groups of block dimensions, and obtain the feature data of the template iris convolution data within the sub-blocks of the i-th, i=1,…,I groups of block dimensions; i represents the index number of the group number of the block dimension; I represents the total number of groups of the block dimension; to obtain the g-th... i -1 sub-block, g-th i Sub-blocks and the g-th i The relationship between the feature data of the template iris convolutional data in +1 sub-blocks;

[0035] Preferably, the feature data of the template iris convolution data within the sub-blocks of the i-th, i=1,...,I-th group block dimension are obtained as follows;

[0036] Specifically, let the block dimension of the i-th group be M. i ×N i The formula used to obtain the feature data of the template iris convolution data within the sub-block of the i-th group block dimension is as follows:

[0037]

[0038] In the formula: Avg i-g Represents the g-th element in the i-th group block dimension. i Feature data of the template iris convolution data within each sub-block; M i N represents the length of the sub-block in the i-th group's block dimension. i F represents the width of the sub-block in the i-th group's block dimension; g (x,y) represents the g-th... i The value of the template iris convolution data at coordinates (x, y) in each sub-block; x represents the g-th value in the i-th sub-block dimension. i The x-coordinate of the template iris convolution data within each sub-block; y represents the g-th digit in the i-th sub-block dimension. i The vertical axis of the template iris convolution data within each sub-block;

[0039] In this example, the template iris convolution data has a dimension of 256×32. Therefore, a total of 40 block dimensions are set, i = 1,...,40, where M... i =2,4,8,16,32,64,128,256,N i =2,4,8,16,32,

[0040] Preferably, in the block dimension of the i-th, i = 1, ..., I group, the g-th i -1 sub-block, g-th i Sub-blocks and the g-th i The relationships between the feature data of the template iris convolution data in +1 sub-blocks include:

[0041] Avg i-(g-1) Greater than Avg i-g The probability of this is denoted as p. i-g ;

[0042] Avg i-(g-1) Less than Avg i-g The probability of j is denoted as j. i-g ;

[0043] When Avg i-(g-1) Greater than Avg i-g At that time, Avg i-g Greater than Avg i-(g+1) The probability of this is denoted as p1. i-g ;

[0044] When Avg i-(g-1) Greater than Avg i-g At that time, Avg i-g Less than Avg i-(g+1) The probability of this is denoted as p2. i-g ;

[0045] When Avg i-(g-1) Less than Avg i-g At that time, Avg i-g Greater than Avg i-(g+1) The probability of this is denoted as j1. i-g ;

[0046] When Avg i-(g-1) Less than Avg i-g At that time, Avg i-g Less than Avg i-(g+1) The probability of this is denoted as j2. i-g ;

[0047] Among them, Avg i-(g-1) Represents the g-th element in the i-th group block dimension. i Feature data of the template iris convolution data within -1 sub-block; Avg i-(g+1) Represents the g-th element in the i-th group block dimension. i Feature data of the iris convolution data within the +1 sub-block template.

[0048] S4: Based on the block dimension of the i-th, i=1,…,I group, the g-th i -1 sub-block, g-th i Sub-blocks and the g-th i The relationship between the feature data of the template iris convolutional data in the +1 sub-blocks is obtained to obtain the analysis results of the information content of the sub-blocks based on the iris under the block dimension of the i, i = 1, ..., I group, so as to obtain the optimal block dimension of the template iris convolutional data, and then determine the feature data of the template iris convolutional data in the sub-blocks under the optimal block dimension.

[0049] Preferably, the formula used to obtain the information content-based analysis results of sub-blocks under the i-th, i=1,…,I group block dimension is as follows:

[0050]

[0051] Where: G i This represents the analysis results of the sub-blocks in the i-th group's block dimension based on information content.

[0052] Specifically, after obtaining the feature data of each template iris convolutional data within a sub-block of each segment dimension, this embodiment can obtain the magnitude relationship of the feature data of each template iris convolutional data in any two different sub-blocks. Furthermore, it can obtain the probability of the relationship between the feature data of all template iris convolutional data in any two different sub-blocks, thereby enabling the determination of Avg in the training samples. i-(g-1) Avg i-g Avg i-(g+1) The relationship between them. Specifically, in this embodiment, Avg is obtained using conventional methods from 1000 training samples. i-(g-1) Avg i-g Avg i-(g+1) The relationships between them include:

[0053] Avg i-(g-1) Greater than Avg i-g The probability p i-g ;

[0054] Avg i-(g-1) Less than Avg i-g The probability j i-g ;

[0055] In Avg i-(g-1) Greater than Avg i-g In the case of Avg i-g Greater than Avg i-(g+1) The probability p1 i-g Avg i-g Less than Avg i-(g+1) The probability p2 i-g ;

[0056] In Avg i-(g-1) Less than Avg i-g In the case of Avg i-g Greater than Avg i-(g+1) The probability j1 i-g Avg i-g Less than Avg i-(g+1) The probability j2 i-g .

[0057] This allows us to obtain the block dimension (M) of the i-th group. i ×N i Under this condition, the analysis results G of all sub-blocks are based on the amount of information. i ;

[0058] Furthermore, the optimal block dimension for obtaining the template iris convolutional data is as follows:

[0059] When the sub-blocks in the i-th group block dimension have the maximum information content analysis results, the i-th group block dimension is the optimal block dimension for template iris convolution data.

[0060] Specifically, in this embodiment, the analysis results of the 40 group blocks are set as: G1~G 40 Select G1 to G 40 The block dimension corresponding to the maximum value is taken as the optimal block dimension for the template iris convolution data.

[0061] In this embodiment, the segmentation dimension of the template iris data is determined by the dimension of the template iris convolution data: the 256×32 dimension template iris convolution data is segmented sequentially according to the 40 segmentation dimensions in Table 1.

[0062] Table 1. 40-group block dimensions

[0063] Serial Number Block Dimension Serial Number Block Dimension Serial Number Block Dimension Serial Number Block Dimension Serial Number Block Dimension 1 256×32 9 128×4 17 32×16 25 16×2 33 4×8 2 256×16 10 128×2 18 32×8 26 8×32 34 4×4 3 256×8 11 64×32 19 32×4 27 8×16 35 4×2 4 256×4 12 64×16 20 32×2 28 8×8 36 2×32 5 256×2 13 64×8 21 16×32 29 8×4 37 2×16 6 128×32 14 64×4 22 16×16 30 8×2 38 2×8 7 128×16 15 64×2 23 16×8 31 4×32 39 2×4 8 128×8 16 32×32 24 16×4 32 4×16 40 2×2

[0064] Specifically, after obtaining 40 groups of template iris convolutional data from template testers, these grouped data are analyzed based on their information content to determine the optimal block dimension for the template iris convolutional data. This allows for the extraction of feature data from the template iris convolutional data within each sub-block at the optimal block dimension. The optimal block dimension for the template iris convolutional data is set to M×N, and the 256×32 dimension template iris convolutional data is divided into Q sub-blocks. Each sub-block contains a certain amount of template iris convolution data;

[0065] S5: Based on the feature data of the template iris convolution data within the sub-blocks under the optimal block dimension, determine whether the g-th sub-block and the g'-th sub-block under the optimal block dimension are stable; where g and g' are the index numbers of the sub-blocks under the optimal block dimension.

[0066] Preferably, the method used to determine whether the g-th sub-block and the g'-th sub-block are in a stable relationship under the optimal partitioning dimension is as follows:

[0067] When, in the g-th sub-block under the optimal partitioning dimension, there exist 75% × X template iris convolutional data whose feature data is greater than the feature data of the corresponding template iris convolutional data in the g'-th sub-block, or 75% × X template iris convolutional data whose feature data is less than the feature data of the corresponding template iris convolutional data in the g'-th sub-block, or 75% × X template iris convolutional data whose feature data is equal to the feature data of the corresponding template iris convolutional data in the g'-th sub-block, the g-th sub-block and the g'-th sub-block under the optimal partitioning dimension are in a stable relationship.

[0068] In this embodiment, 1000 images were selected as training samples. Therefore, in this embodiment, when there are 750 template iris convolutional data features in the g-th sub-block under the optimal block dimension that are greater than the feature data of the corresponding template iris convolutional data in the g'-th sub-block, or when there are 750 template iris convolutional data features that are less than the feature data of the corresponding template iris convolutional data in the g'-th sub-block, or when there are 750 template iris convolutional data features that are equal to the feature data of the corresponding template iris convolutional data in the g'-th sub-block, the g-th sub-block and the g'-th sub-block under the optimal block dimension are in a stable relationship.

[0069] Specifically, after determining the optimal partitioning dimension of the template, the number of sub-blocks under the optimal partitioning dimension can be obtained, and their total number is... Then, the average value of the template iris convolution data of each of the Q sub-blocks is calculated as the feature data of that sub-block. The relative relationship between the feature data of each sub-block and other sub-blocks is statistically analyzed under the existing 1000 training samples. Q represents the total number of sub-blocks under the optimal block dimension; M represents the length of the sub-block under the optimal block dimension; and N represents the width of the sub-block under the optimal block dimension.

[0070] If, under the optimal partitioning dimension, there are 750 template iris convolutional data features in the g-th sub-block that are greater than, less than, or equal to the corresponding template iris convolutional data in the g'-th sub-block, then the relative relationship between the g-th and g'-th sub-blocks under the optimal partitioning dimension is considered stable. Otherwise, the relative relationship between the g-th and g'-th sub-blocks under the optimal partitioning dimension is considered unstable. The stable relationships between each sub-block and other sub-blocks under the optimal partitioning dimension are determined sequentially to form a stable relationship set.

[0071] S6: When the g-th sub-block and the g'-th sub-block are in a stable relationship under the optimal partitioning dimension, determine whether the g-th sub-block and the g'-th sub-block are in a constant relationship under the optimal partitioning dimension; in order to establish a set of constant relationships;

[0072] Preferably, the method for determining whether the g-th sub-block and the g'-th sub-block have a constant relationship under the optimal partitioning dimension is as follows:

[0073] Specifically, after determining the stable set of relationships in the template tester feature data, it is then determined whether these stable relationships are constant relationships:

[0074] S61: First, obtain the constant parameters between the g-th sub-block and the g'-th sub-block under the optimal partitioning dimension, as shown in the following formula:

[0075]

[0076] In the formula: P Constant λ is a constant parameter; yb is the index of the training sample; YB is the total number of training samples; Sgn(·) represents the sign function; e is the natural base; λ is a variable empirical parameter value; k represents the total number of template iris convolution data in the g-th sub-block of the yb-th training sample; o represents the index number of the template iris convolution data in the g-th sub-block of the yb-th training sample under the optimal block dimension; u g Q is the average of the feature data of the g-th sub-block under the optimal block dimension; g' This represents the feature data of the g'th sub-block under the optimal block dimension; The threshold for comparing the feature data of the g'th sub-block under the optimal block dimension; denoted as the standard deviation of the feature data of the g'th sub-block under the optimal block dimension;

[0077] S62: When the constant parameter is greater than or equal to the set constant judgment threshold, the g-th sub-block and the g'-th sub-block under the optimal block dimension are in a constant relationship;

[0078] Specifically, a constant judgment threshold PC is set; if P Constant If ≥PC, then the g-th sub-block and the g'-th sub-block under the optimal partitioning dimension are considered to be constant relations, and the final constant relation set of the template tester can be obtained. Here, d is the index number of the constant relation, and D is the total number of constant relations;

[0079] S7: When the g-th sub-block and the g'-th sub-block have a constant relationship under the optimal block segmentation dimension, obtain the complex features of the sub-blocks that have a constant relationship, and then obtain the complex features of the template testers;

[0080] Preferably, the method for obtaining the complex features of the template tester is as follows:

[0081] S71: Obtain the distance between the g-th sub-block and the g-th sub-block that has a constant relationship with the g-th sub-block under the optimal partitioning dimension, and sort the sub-blocks that have a constant relationship with the g-th sub-block in ascending order of distance to obtain the sorted set of sub-blocks that have a constant relationship with the g-th sub-block; where the first element in the set has the smallest distance to the g-th sub-block under the optimal partitioning dimension.

[0082] S72: Obtain the real and imaginary parts of the g-th sub-block under the optimal block dimension to obtain the complex features of the g-th sub-block under the optimal block dimension;

[0083] Preferably, in S72, the real part of the g-th sub-block under the optimal partitioning dimension is:

[0084] R g-1 =Q g (4) The real and imaginary parts of the g-th sub-block under the optimal block dimension are:

[0085]

[0086] In the formula: R g-2 Y1 represents the imaginary part of the g-th sub-block under the optimal block dimension; Y1 represents the feature data of the first sub-block in the set of sorted sub-blocks that have a constant relationship with the g-th sub-block, where the feature data of each sub-block is the average value of all template iris convolution data in that sub-block; Q represents the feature data of the first sub-block in the set of sorted sub-blocks that have a constant relationship with the g-th sub-block. g Let represent the feature data of the g-th sub-block under the optimal partitioning dimension; u represents the distance between the 1-th sub-block and the g-th sub-block in the sorted set of sub-blocks that have a constant relationship with the g-th sub-block; |·| represents the absolute value; R g-1 Let be the real part of the g-th sub-block under the optimal block dimension;

[0087] Then, the complex feature of the g-th sub-block under the optimal block dimension is obtained as R. g-1 +R g-2 ×r, where r represents the imaginary unit.

[0088] Specifically, the complex feature of the g-th sub-block of a single training sample is given by the real part R. g-1 The feature data representing the g-th sub-block, with the imaginary part R g-2 The angle value represents the change in feature data between the g-th sub-block and the feature data of other sub-blocks.

[0089] Specifically, in this embodiment, based on the real and imaginary parts of the complex features of a single sample, the data distribution of 1000 training samples is analyzed, and the simulation function of the real and imaginary parts is fitted by the least squares method as the template for the complex features of the test personnel.

[0090] S73: Obtain the representation of the real part of the complex feature of the template tester in the g-th sub-block under the optimal block dimension;

[0091] Preferably, the method for obtaining the representation of the real part of the complex feature of the template tester in the g-th sub-block under the optimal block dimension is as follows:

[0092] When the first sub-block in the sorted set of sub-blocks having a constant relationship with the g-th sub-block and the second sub-block in the sorted set of sub-blocks having a constant relationship with the g-th sub-block are respectively located on both sides of the g-th sub-block under the optimal block dimension,

[0093] If Y1 ≤ Q g ≤ Y2, or Y2 ≤ Q g ≤ Y1, then the representation of the real part of the complex feature of the template tester in the g-th sub-block under the optimal block dimension is a linear function of one variable; at this time, it shows that Q g , Y1 and Y2 have a linear variation relationship, so the real part of the complex feature of the template tester in the g-th sub-block under the optimal block dimension adopts the form of a linear function of one variable a1×x + b1;

[0094] If Q g < min(Y1, Y2), or Q g > max(Y1, Y2), then the representation of the real part of the complex feature of the template tester in the g-th sub-block under the optimal block dimension is a quadratic function of one variable; at this time, Q g is smaller than the minimum value min(Y1, Y2) of Y1 and Y2, or Q g is larger than the maximum value max(Y1, Y2) of Y1 and Y2, which shows that Q g , Y1 and Y2 have a curve variation relationship, so the real part of the complex feature of the template tester adopts the form of a quadratic function of one variable a2×x 2 + b2×x + c;

[0095] If the first sub-block in the sorted set of sub-blocks having a constant relationship with the g-th sub-block, the second sub-block in the sorted set of sub-blocks having a constant relationship with the g-th sub-block, and the g-th sub-block under the optimal block dimension are adjacent in sequence and satisfy Y1 < Y2 < Q g , or the g-th sub-block under the optimal block dimension, the first sub-block in the sorted set of sub-blocks having a constant relationship with the g-th sub-block, and the second sub-block in the sorted set of sub-blocks having a constant relationship with the g-th sub-block are adjacent in sequence and satisfy Q g < Y1 < Y2, then the representation of the real part of the complex feature of the template tester in the g-th sub-block under the optimal block dimension is a linear function of one variable; at this time, it shows that Q g, Y1 and Y2 show a linear variation relationship. Therefore, the real part of the complex number feature of the template tester adopts the form of a linear function a1×x + b1;

[0096] If the first sub-block in the set of sub-blocks having a constant relationship with the g-th sub-block after sorting, the second sub-block in the set of sub-blocks having a constant relationship with the g-th sub-block after sorting, and the g-th sub-block under the optimal block dimension are adjacent in sequence, and do not satisfy Y1 < Y2 < Q g , or the g-th sub-block under the optimal block dimension, the first sub-block in the set of sub-blocks having a constant relationship with the g-th sub-block after sorting, and the second sub-block in the set of sub-blocks having a constant relationship with the g-th sub-block after sorting are adjacent in sequence, and do not satisfy Q g < Y1 < Y2, then the representation form of the real part of the complex number feature of the template tester in the g-th sub-block under the optimal block dimension is a quadratic function. At this time, the real part of the complex number feature of the template tester adopts the quadratic function a2×x 2 + b2×x + c;

[0097] Specifically, in this embodiment, select the two sub-blocks Y1 and Y2 that have a constant relationship with the g-th sub-block Q g and are the closest to the g-th sub-block Q g . Determine the representation form of the real part of the complex number feature of the template tester in the g-th sub-block under the optimal block dimension according to the mutual relationship of Y1, Y2, and Q g .

[0098] S74: According to the complex number feature of the g-th sub-block under the optimal block dimension (i.e., the real part R g-1 ) of the g-th sub-block under the optimal block dimension, and the representation form of the real part of the complex number feature of the template tester in the g-th sub-block under the optimal block dimension, use the least squares method to fit the template iris convolution data corresponding to X training data, and obtain the real part L g-1 of the complex number feature of the template tester in the g-th sub-block under the optimal block dimension;

[0099] Specifically, after determining the representation form of the real part of the complex number feature of the template tester in the g-th sub-block under the optimal block dimension, fit the feature data of 1000 training samples in the g-th sub-block in the least squares method, and the function obtained by fitting R g-1 is used as the real part of the complex number feature of the template tester;

[0100] S75: Obtain the imaginary part of the complex number feature of the template tester as: L g-2 = k1×L g-1 + k2.

[0101] Specifically: k1, k2 are for calculating Lg-2 The function parameters at time, Rg is fitted to the complex features of the g-th sub-block of 1000 training samples using the least squares method. g-1 , to obtain L g-1 Subsequently, the imaginary part L of the complex features of the template tester is then used. g-2 ;

[0102] Specifically, all template testers are processed according to steps S1 to S7 to obtain the complex characteristics of all template testers;

[0103] S8: Obtain the grayscale image of the test iris of the tester to obtain the normalized enhanced image of the test iris, and then obtain the convolutional data of the test iris;

[0104] Specifically, after obtaining the complex features of each template tester, an iris grayscale image of the tester's test iris is acquired using an iris scanner. The computer system converts the test iris grayscale image into a 256×32 dimensional test iris normalized and enhanced image using the Daugman rubber band method and equalized histogram. The test iris normalized and enhanced image of the tester is then convolved using a gradient Laplacian convolution kernel. The grayscale value of the convolved test iris normalized and enhanced image is read to obtain 256×32 dimensional test iris convolution data.

[0105] S9: Based on the optimal segmentation dimension, obtain the test iris feature data of the g-th sub-block; and based on the constant relation set, determine whether the test iris feature data of the g-th sub-block may belong to the template tester.

[0106] Specifically, based on the optimal block dimension, the mean of the test iris convolution data of the g-th sub-block is obtained as the test iris feature data of the g-th sub-block; it is then determined whether the test iris feature data of the g-th sub-block may belong to the template tester.

[0107] Preferably, the method used to determine whether the test iris feature data of the g-th sub-block may belong to the template tester is as follows:

[0108] Specifically, the test iris convolution data is segmented according to the template tester using the optimal block dimension, and the average value of a single sub-block is read as the test iris feature data y1~y2. Q ,

[0109] If there exists a set of constant relational data containing test iris feature data of a sub-block that has a constant relation with the g-th sub-block, and the relationship between the test iris feature data of the g-th sub-block and the test iris feature data of the g-th sub-block is not constant, then the test iris feature data of the g-th sub-block cannot possibly belong to the template tester; in this case, the tester is not a template tester, and no further verification of the tester is required.

[0110] If the test iris feature data of the sub-block with a constant relationship to the g-th sub-block in the constant relation set has a constant relationship with the test iris feature data of the g-th sub-block, then the test iris feature data of the g-th sub-block may belong to the template tester; in this case, the test iris data is verified a second time.

[0111] S10: When the test iris feature data of the g-th sub-block may belong to the template tester, obtain the real part representation of the complex feature of the test iris feature data of the g-th sub-block, and obtain the complex feature of the tester based on the least squares method.

[0112] Specifically, after qualitative analysis, the number of template testers was reduced, resulting in a new range of template testers; and based on the complex features of the g-th sub-block under the optimal partitioning dimension, the testers' feature data were fitted using the least squares method to obtain the testers' feature data y1~y q Complex characteristics for each template tester;

[0113] S11: Determine whether a tester belongs to the template tester category based on the multiple characteristics of template testers and multiple characteristics of testers.

[0114] Specifically, the complex features of the tester are matched and compared with the complex features of the corresponding template tester by subtraction and modulo to obtain the final identification conclusion;

[0115] Specifically, the real part of the complex feature of the tester in the g-th sub-block is set to be... The imaginary part is

[0116] The complex characteristics of the template tester are obtained in step S7, L g-1 The real part is L g-1 The imaginary part is L g-2 ;

[0117] S111: If the real part of the tester and the real part of the template tester have the same representation in the g-th sub-block, then set the exponential logic judgment value H(g) of the g-th sub-block to 1.

[0118] If the real part of the tester and the real part of the template tester are represented differently in the g-th sub-block, then the exponential logic judgment value H(g) of the g-th sub-block is set to 0.

[0119] S112: Obtain the matching modulus distance LC between the tester and the template tester; to determine the template tester with the smallest matching modulus distance to the tester;

[0120]

[0121] In the formula: G represents the total number of sub-blocks under the optimal partitioning dimension; Let be the real part of the complex feature of the tester in the g-th sub-block; L represents the imaginary part of the complex feature of the tester in the g-th sub-block; g-1 L is the real part of the complex feature of the template tester in the g-th sub-block; g-2 Let be the imaginary part of the complex features of the template testers in the g-th sub-block;

[0122] Specifically, the complex features of the tester's iris are compared with the complex features of each template tester to determine the template tester with the smallest LC matching modulus distance.

[0123] S113: If the template tester with the smallest matching modulus distance to the tester and the tester's exponential logic judgment value H(g) on ​​all sub-blocks are both equal to 1, and the matching modulus distance is less than the matching threshold μ, then the tester is determined to belong to the template tester with the smallest matching modulus distance to the tester, and the tester sample is considered to match the template tester sample; otherwise, the tester is considered not to match any template testers. Finally, the final recognition result is fed back to the tester as the recognition conclusion.

[0124] Example 1:

[0125] The entire process of the operation on five people (named B1, B2, B3, B4, and B5, whose information had not been entered before; identity verification tests were conducted on B1, B2, B3, B4, and B5; the test irises and template irises were collected using the same iris scanner):

[0126] 1) Using any commercially available iris scanner, collect 1000 grayscale images of the template iris for each of the experimenters B1, B2, B3, B4, and B5.

[0127] 2) The computer system uses the Daugman rubber band method and the equalization histogram method to convert 1,000 template iris grayscale images of B1, B2, B3, B4, and B5 into 256×32 dimension template iris normalization and enhancement images respectively.

[0128] 3) Perform convolution processing on all template iris normalization enhancement images of B1, B2, B3, B4, and B5 using gradient Laplacian convolution kernels. Read the grayscale value of each template iris normalization enhancement image after convolution to obtain the 256×32 dimension template iris convolution data of each template iris normalization enhancement image of B1, B2, B3, B4, and B5.

[0129] 4) The 256×32 dimension template iris convolutional data of B1, B2, B3, B4, and B5 were segmented into blocks using the 40 block dimensions in Table 1. Based on information content analysis, the optimal block dimension for B1 was determined to be 32×4, for B2 16×32, for B3 64×8, for B4 16×4, and for B5 8×16. Therefore, the number of sub-blocks for B1 was determined to be 64, for B2 16, for B3 16, for B4 64, and for B5 64.

[0130] 5) Determine the stable relationships between sub-blocks B1, B2, B3, B4, and B5 respectively, where:

[0131] The set of stable relations for sub-blocks of B1 is:

[0132] {1,3,8,10,15,16,17,18,19,22,25,26,27,28,29,30,31,32,33,38,41,42,43,44,45,46,47,48,49,52,53,54,56,58,60,61,62}

[0133] The set of stable relations for sub-blocks of B2 is:

[0134] {1,2,3,4,5,7,8,10,11,12,14,15}

[0135] The set of stable relations for sub-blocks of B3 is as follows:

[0136] {2,3,4,5,6,7,8,9,10,14,15,16}

[0137] The set of stable relations for the sub-blocks of B4 is:

[0138] {3,4,5,6,7,8,13,14,15,19,20,21,22,23,24,28,29,30,31,32,33,34,35,36,41,42,45,46,47,48,49,53,54,55,56,57,58,61,62,63,64}

[0139] The set of stable relations for the sub-blocks of B5 is as follows:

[0140] {1,6,8,9,10,11,12,13,14,18,20,21,22,25,26,27,28,29,30,35,36,37,38,39,44,45,46,47,50,51,52,53,55,56,58,62,63,64}

[0141] The stable relationships of B1, B2, B3, B4, and B5 are analyzed separately to determine the constant relationships, where:

[0142] The constant relationship of B1 is:

[0143] {1,3,8,10,15,25,26,30,31,32,33,38,41,46,47,48,49,56,58}

[0144] The constant relationship of B2 is:

[0145] {2,3,4,5,10,14,15,16}

[0146] The constant relationship of B3 is:

[0147] {2,3,7,8,9,10,14}

[0148] The constant relationship for B4 is:

[0149] {13,14,15,19,20,28,29,30,31,32,33,34,35,36,41,42,45,46,47,48,49,53,54}

[0150] The constant relationship of B5 is:

[0151] {1,6,8,9,22,25,26,27,28,29,30,35,39,44,45,46,47,50,51,52,53,55}

[0152] 6) Based on the constant relationships of B1, B2, B3, B4, and B5, determine the complex characteristics of B1, B2, B3, B4, and B5 respectively.

[0153] 7) Collect one grayscale image of the test iris of B1 using an iris scanner.

[0154] 8) The computer system uses the Daugman rubber band method and equalization histogram to convert the B1 test iris grayscale image into a 256×32 dimension test iris normalized enhancement image.

[0155] 9) The test iris normalization enhancement image of B1 is convolved using the same gradient Laplacian convolution kernel as in 3). The grayscale value of the convolved image of each template iris normalization enhancement image is read to obtain the 256×32 dimension test iris convolution data of the test iris normalization enhancement image of B1.

[0156] 10) The test iris convolution data of B1 is divided into the optimal block dimensions of B1, B2, B3, B4, and B5 respectively. The feature data of the test iris is read when matching. By performing a qualitative analysis on the constant relationship between the test iris convolution data of B1 and B1, B2, B3, B4, and B5, B3 and B2 are eliminated.

[0157] 11) The test iris convolution data of B1 are fitted with complex features based on the constant relationship between B1, B4 and B5 respectively.

[0158] 12) Compare the complex features of the test iris of B1 with the complex features of B1, B4, and B5 respectively. B4 has an exponential logic judgment value of 0. The matching modulus distance LC of B5 is 0.685, and the matching modulus distance LC of B1 is 0.165. Set the matching threshold μ = 0.4. B1 has the smallest matching modulus distance, and B1's LC < μ. Therefore, the conclusion is that the test image belongs to the same category as B1. Thus, the authentication result is output as B1.

[0159] Example 2:

[0160] The entire process of the operation on six people (named C1, C2, C3, C4, C5, and C6; information of experimenters C1, C2, C3, C4, and C5 had been previously entered; identity verification tests were performed on experimenters C1, C2, C3, C4, C5, and C6; the test irises and template irises were collected using the same iris scanner):

[0161] 1) Acquire one grayscale image of the test iris of C6 using an iris scanner.

[0162] 2) The computer system uses the Daugman rubber band method and equalization histogram to convert the grayscale image of the test iris of C6 into a 256×32 dimension test iris normalized enhancement image.

[0163] 3) The test iris normalization enhancement image of C6 is convolved using a gradient Laplacian convolution kernel. The grayscale value of the convolved image of each template iris normalization enhancement image is read to obtain the 256×32 dimension test iris convolution data of the test iris normalization enhancement image of C6.

[0164] 4) The test iris convolution data of C6 were segmented using the optimal block dimensions of C1, C2, C3, C4, and C5, respectively. The feature data of the test iris was read for matching. By performing a qualitative analysis on the constant relationship between the test iris convolution data of C6 and C1, C2, C3, C4, and C5, the result is that when comparing C6, C2, C3, and C5 were eliminated.

[0165] 5) The test iris convolution data for C6 were fitted with complex features based on the constant relationship between C1 and C4, respectively;

[0166] 6) Compare the complex features of the test iris of C6 with those of C1 and C4 respectively. The matching modulus distance LC of C1 is 0.758, and that of C4 is 0.555. Set the matching threshold μ = 0.4. C4 has the smallest matching modulus distance, but C4's LC > μ. Therefore, we conclude that the test image of C6 does not match any of the template testers, and the final recognition result is output.

[0167] Example 3:

[0168] The entire process of the operation on two people (named A1 and A2, with information of experimenter A1 previously entered; identity verification tests were performed on experimenters A1 and A2, and the iris scans of the test iris and the template iris were collected using the same iris scanner):

[0169] 1) Collect one grayscale image of the test iris of A2 using an iris scanner.

[0170] 2) The computer system uses the Daugman rubber band method and equalization histogram to convert the grayscale image of the test iris of A2 into a 256×32 dimension test iris normalized enhancement image.

[0171] 3) The test iris normalization enhancement image of A2 is convolved using a gradient Laplacian convolution kernel. The grayscale value of each template iris normalization enhancement image after convolution is read to obtain the 256×32 dimension test iris convolution data of the test iris normalization enhancement image of A2.

[0172] 4) The test iris convolutional data of A2 is segmented using the optimal block dimension of A1. The feature data of the test iris is read for matching. By performing a qualitative analysis on the constant relationship between the test iris convolutional data of A2 and A1, it is found that the constant relationship between A2 and A1 does not match. Therefore, it is determined that A2 does not belong to A1, and the recognition conclusion is output.

[0173] from Figure 3 and Figure 4As can be seen from the figure, the verification method of this embodiment has a significant advantage in recognition accuracy compared with existing methods: in the JLU and CASIA iris databases, the EER values ​​of the ROC curves are 0.09% and 0.12% respectively, which is 25%-30% lower than existing methods such as generative adversarial networks and end-to-end frameworks. In the figure, Case 0: the method of this embodiment; Case 1: iris recognition using generative adversarial networks; Case 2: accurate iris segmentation and recognition based on the end-to-end unified framework of MADNet and DSANet; Case 3: a hidden iris recognition method; Case 4: an iris recognition algorithm based on sequence metric neural networks; Case 5: an iris feature recognition method based on k-means clustering analysis.

[0174] The verification method in this embodiment innovatively integrates texture region segmentation with complex spatial feature representation to construct a two-dimensional recognition model based on "dynamic texture-angle change," breaking through the dependence of traditional methods on high-performance computing resources. Its technical advantages are not only reflected in the balance between recognition accuracy and efficiency, but also in providing a new path for the popularization of iris recognition technology in low-end devices through "algorithm optimization replacing hardware upgrades," making it particularly suitable for identity verification scenarios sensitive to power consumption and cost.

[0175] This embodiment employs a texture region-guided segmentation strategy based on information content to accurately select the segmentation scheme that best reflects the iris texture features, effectively overcoming the limitations of traditional methods that directly describe texture changes to obtain iris information. By subdividing the iris image through optimal block dimensions, it is possible to deeply mine local texture features, enhance sensitivity to iris texture changes, and lay a data foundation for accurate recognition. Simultaneously, this strategy significantly reduces data redundancy and lowers the model's computational resource requirements, enabling the iris recognition model to run efficiently on resource-constrained, low-computing-power devices, reducing dependence on high-end hardware and expanding the practical application scope and feasibility of the technology. This improves the scientific rigor and accuracy of iris texture information analysis, reduces data redundancy, and lowers the demand for computational resources.

[0176] This embodiment extends the traditional real-number domain feature representation to the complex space. Each dimension's complex representation includes a real part (dimensional data) and an imaginary part (changing angle). This dual-dimensional representation essentially increases the information capacity of a single dimension, enabling a more nuanced and comprehensive depiction of the complex variations in iris texture. It overcomes the inherent limitations of real-number domain representation, improving the richness and diversity of features. By fitting the real and imaginary part functions of the complex features using the least squares method, the feature representation is mathematically optimized, enhancing its accuracy and robustness. This provides mathematical support for the accuracy of iris recognition, directly improving the accuracy and reliability of recognition, and significantly enhancing the application effect of iris recognition technology in real-world scenarios. Expanding the feature representation dimension enhances the richness, diversity, accuracy, and robustness of features, thereby improving recognition accuracy and reliability.

[0177] This embodiment organically combines texture region variation patterns with complex space feature representation to form a comprehensive strategy that effectively addresses the uncertainties in iris image acquisition. It ensures a constant relative relationship between dimensions even when iris image data fluctuates, providing a stable theoretical framework for feature matching under uncertain conditions and guaranteeing the stability and reliability of feature matching from a scientific perspective. The utilization of complex space not only enriches feature representation but also improves recognition efficiency and accuracy through the characteristics of complex number operations. This allows the model to maintain high recognition performance even when processing complex and variable iris images, enhancing the system's practicality and adaptability, and providing a reliable solution for iris recognition in resource-constrained environments. It effectively addresses the uncertainties in iris image acquisition, enhancing the system's practicality and adaptability.

[0178] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A low-computing-power authentication method for scenarios involving iris information content and complex field features, characterized in that, Includes the following steps: S1: Obtain X template iris grayscale images of multiple template testers as training data, and normalize the template iris grayscale images to obtain template iris normalized enhanced images; where X represents the total number of template iris grayscale images used as training data. S2: Based on the template iris normalization enhancement image, obtain template iris convolution data using gradient Laplacian convolution kernels; S3: Randomly set the I-th group block dimension, obtain the feature data of the template iris convolution data in the sub-blocks under the i-th, i=1,…,I-th group block dimension, so as to obtain the g-th... i -1 sub-block, g-th i Sub-blocks and the g-th i The relationships between the feature data of the template iris convolutional data in +1 sub-blocks; where i represents the index number of the group in the block dimension; I represents the total number of groups in the block dimension; g i This represents the index number of the sub-block under the i-th group's block dimension; S4: Based on the block dimension of the i-th, i=1,…,I group, the g-th i -1 sub-block, g-th i Sub-blocks and the g-th i The relationship between the feature data of the template iris convolutional data in the +1 sub-blocks is obtained to obtain the analysis results of the information content of the sub-blocks based on the iris under the block dimension of the i, i = 1, ..., I group, so as to obtain the optimal block dimension of the template iris convolutional data, and then determine the feature data of the template iris convolutional data in the sub-blocks under the optimal block dimension. S5: Based on the feature data of the template iris convolution data within the sub-blocks under the optimal block dimension, determine whether the g-th sub-block and the g'-th sub-block under the optimal block dimension are stable; where g and g' are the index numbers of the sub-blocks under the optimal block dimension. S6: When the g-th sub-block and the g'-th sub-block are stable under the optimal partitioning dimension, determine whether the g-th sub-block and the g'-th sub-block are constant under the optimal partitioning dimension, so as to establish a constant relationship set; S7: When the g-th sub-block and the g'-th sub-block have a constant relationship under the optimal block segmentation dimension, obtain the complex features of the sub-blocks that have a constant relationship, and then obtain the complex features of the template testers; S8: Obtain the grayscale image of the test iris of the tester to obtain the normalized enhanced image of the test iris, and then obtain the convolutional data of the test iris; S9: Based on the optimal segmentation dimension, obtain the test iris feature data of the g-th sub-block; to determine whether the test iris feature data of the g-th sub-block may belong to the template tester; S10: When the test iris feature data of the g-th sub-block may belong to the template tester, obtain the real part representation of the complex feature of the test iris feature data of the g-th sub-block to obtain the complex feature of the tester; S11: Based on the multiple characteristics of template testers and multiple characteristics of testers, determine whether a tester belongs to the template tester, and realize identity verification based on iris information.

2. The low-computing-power scenario identity verification method based on iris information content and complex field features according to claim 1, characterized in that, In S3, the feature data of the template iris convolution data in the sub-blocks under the block dimension of the i,i=1,…,I group are as follows; In the formula: Avg i-g Represents the g-th element in the i-th group block dimension. i Feature data of template iris convolution data within each sub-block; M i N represents the length of the sub-block in the i-th group's block dimension. i F represents the width of the sub-block in the i-th group's block dimension; g (x,y) represents the g-th... i The value of the template iris convolution data at coordinates (x, y) in each sub-block; x represents the g-th value in the i-th sub-block dimension. i The x-coordinate of the template iris convolution data within each sub-block; y represents the g-th digit in the i-th sub-block dimension. i The vertical axis of the template iris convolution data within each sub-block.

3. The low-computing-power scenario identity verification method based on iris information content and complex field features according to claim 2, characterized in that, In S3, under the block dimension of the i-th, i=1,…,I group, the g-th i -1 sub-block, g-th i Sub-blocks and the g-th i The relationships between the feature data of the template iris convolution data in +1 sub-blocks include: Avg i-(g-1) Greater than Avg i-g The probability of this is denoted as p. i-g ; Avg i-(g-1) Less than Avg i-g The probability of j is denoted as j. i-g ; When Avg i-(g-1) Greater than Avg i-g At that time, Avg i-g Greater than Avg i-(g+1) The probability of this is denoted as p1. i-g ; When Avg i-(g-1) Greater than Avg i-g At that time, Avg i-g Less than Avg i-(g+1) The probability of this is denoted as p2. i-g ; When Avg i-(g-1) Less than Avg i-g At that time, Avg i-g Greater than Avg i-(g+1) The probability of this is denoted as j1. i-g ; When Avg i-(g-1) Less than Avg i-g At that time, Avg i-g Less than Avg i-(g+1) The probability of this is denoted as j2. i-g ; Among them, Avg i-(g-1) Represents the g-th element in the i-th group block dimension. i Feature data of the template iris convolution data within -1 sub-block; Avg i-(g+1) Represents the g-th element in the i-th group block dimension. i Feature data of the iris convolution data within the +1 sub-block template.

4. The low-computing-power scenario identity verification method based on iris information content and complex field features according to claim 3, characterized in that, In S4, The formula used to obtain the information content-based analysis results of sub-blocks under the i-th, i=1,…,I group block dimension is as follows: In the formula: G i This represents the analysis results of the sub-blocks in the i-th group's block dimension based on information content; The optimal block dimension for obtaining the template iris convolution data is as follows: When the sub-blocks in the i-th group block dimension have the maximum information content analysis results, the i-th group block dimension is the optimal block dimension for template iris convolution data.

5. The low-computing-power scenario authentication method based on iris information content and complex field features according to claim 4, characterized in that, In S5, the method used to determine whether the g-th sub-block and the g'-th sub-block are in a stable relationship under the optimal partitioning dimension is as follows: When, in the g-th sub-block under the optimal partitioning dimension, there exist 75% × X template iris convolutional data features that are greater than the corresponding template iris convolutional data features in the g'-th sub-block, or 75% × X template iris convolutional data features that are less than the corresponding template iris convolutional data features in the g'-th sub-block, or 75% × X template iris convolutional data features that are equal to the corresponding template iris convolutional data features in the g'-th sub-block, the g-th sub-block and the g'-th sub-block under the optimal partitioning dimension are in a stable relationship.

6. The low-computing-power scenario identity verification method based on iris information content and complex field features according to claim 5, characterized in that, In step S6, the method for determining whether the g-th sub-block and the g'-th sub-block have a constant relationship under the optimal block dimension is as follows: S61: First, obtain the constant parameters between the g-th sub-block and the g'-th sub-block under the optimal partitioning dimension, as shown in the following formula: In the formula: P Constant λ is a constant parameter; yb is the index of the training sample; YB is the total number of training samples; Sgn(·) represents the sign function; e is the natural base; λ is a variable empirical parameter value; k represents the total number of template iris convolution data in the g-th sub-block of the yb-th training sample; o represents the index number of the template iris convolution data in the g-th sub-block of the yb-th training sample under the optimal block dimension; u g Q is the average of the feature data of the g-th sub-block under the optimal block dimension; g' This represents the feature data of the g-th sub-block under the optimal block dimension; The threshold for comparing the feature data of the g'th sub-block under the optimal block dimension; denoted as the standard deviation of the feature data of the g'th sub-block under the optimal block dimension; S62: When the constant parameter is greater than or equal to the set constant judgment threshold, the g-th sub-block and the g'-th sub-block under the optimal block dimension are in a constant relationship.

7. The low-computing-power scenario identity verification method based on iris information content and complex field features according to claim 6, characterized in that, The method for obtaining the complex characteristics of template testers is as follows: S71: Obtain the distance between the g-th sub-block and the g-th sub-block that has a constant relationship with the g-th sub-block under the optimal partitioning dimension, and sort the sub-blocks that have a constant relationship with the g-th sub-block in ascending order of distance to obtain the sorted set of sub-blocks that have a constant relationship with the g-th sub-block. S72: Obtain the real and imaginary parts of the g-th sub-block under the optimal block dimension to obtain the complex features of the g-th sub-block under the optimal block dimension; In S72, the real part of the g-th sub-block under the optimal block dimension is: R g-1 =Q g In S72, the real part of the g-th sub-block under the optimal block dimension is: In the formula: R g-2 Y is the imaginary part of the g-th sub-block under the optimal block dimension; Y1 is the feature data of the first sub-block in the set of sorted sub-blocks that have a constant relationship with the g-th sub-block; Q g Let represent the feature data of the g-th sub-block under the optimal partitioning dimension; u represents the distance between the 1-th sub-block and the g-th sub-block in the sorted set of sub-blocks that have a constant relationship with the g-th sub-block; |·| represents the absolute value; R g-1 Let be the real part of the g-th sub-block under the optimal block dimension; Then, the complex feature of the g-th sub-block under the optimal block dimension is obtained as R. g-1 +R g-2 ×r, where r represents the imaginary unit; S73: Obtain the representation of the real part of the complex features of the template tester in the g-th sub-block under the optimal block dimension; The method for obtaining the representation of the real part of the complex features of the template tester in the g-th sub-block under the optimal block dimension is as follows: When the first sub-block in the sorted set of sub-blocks with a constant relationship to the g-th sub-block and the second sub-block in the sorted set of sub-blocks with a constant relationship to the g-th sub-block are located on either side of the g-th sub-block in the optimal partitioning dimension, If Y1≤Q g ≤Y2, or Y2≤Q g If ≤Y1, then the real part of the complex feature of the template tester in the g-th sub-block under the optimal block dimension is represented as a linear function in one variable; If Q g < min(Y1, Y2), or Q g > max(Y1, Y2), then the representation of the real part of the complex feature of the template tester in the g-th sub-block under the optimal block dimension is a quadratic function of one variable; If the first sub-block in the set of sub-blocks having a constant relationship with the g-th sub-block after sorting, the second sub-block in the set of sub-blocks having a constant relationship with the g-th sub-block after sorting, and the g-th sub-block under the optimal block dimension are adjacent in sequence, and satisfy Y1 < Y2 < Q g , or the g-th sub-block under the optimal block dimension, the first sub-block in the set of sub-blocks having a constant relationship with the g-th sub-block after sorting, and the second sub-block in the set of sub-blocks having a constant relationship with the g-th sub-block after sorting are adjacent in sequence, and satisfy Q g < Y1 < Y2, then the representation form of the real part of the complex number feature of the template tester in the g-th sub-block under the optimal block dimension is a linear function; If the first sub-block in the set of sub-blocks having a constant relationship with the g-th sub-block after sorting, the second sub-block in the set of sub-blocks having a constant relationship with the g-th sub-block after sorting, and the g-th sub-block under the optimal block dimension are adjacent in sequence and do not satisfy Y1 < Y2 < Q g , or the g-th sub-block under the optimal block dimension, the first sub-block in the set of sub-blocks having a constant relationship with the g-th sub-block after sorting, and the second sub-block in the set of sub-blocks having a constant relationship with the g-th sub-block after sorting are adjacent in sequence and do not satisfy Q g < Y1 < Y2, then the representation form of the real part of the complex characteristics of the template tester in the g-th sub-block under the optimal block dimension is a quadratic function of one variable; S74: Based on the representation of the real part of the complex feature of the g-th sub-block under the optimal block partitioning dimension and the real part of the complex feature of the template tester in the g-th sub-block under the optimal block partitioning dimension, the real part L of the complex feature of the template tester in the g-th sub-block under the optimal block partitioning dimension is obtained by using the least squares method. g-1 ; S75: Obtain the imaginary part of the complex features of the template tester to obtain the complex features of the complex features of the template tester; The imaginary part of the complex features of the template tester is: L g-2 =k1×L g-1 +k2; Among them, L g-1 L is the real part of the complex feature of the template tester in the g-th sub-block; g-2 Let k1 and k2 be the imaginary part of the complex features of the template testers in the g-th sub-block; k1 and k2 are both used to calculate L. g-2 Function parameters at that time.

8. The low-computing-power scenario authentication method based on iris information content and complex field features according to claim 7, characterized in that, In step S9, the method used to determine whether the test iris feature data of the g-th sub-block might belong to the template tester is as follows: If there exists a set of constant relational data for the test iris feature data of a sub-block that has a constant relation with the g-th sub-block, and the relationship between the test iris feature data of the g-th sub-block and the test iris feature data of the g-th sub-block is not a constant relation, then the test iris feature data of the g-th sub-block does not belong to the template tester. If the test iris feature data of the sub-block with a constant relationship to the g-th sub-block in the constant relation set has a constant relationship with the test iris feature data of the g-th sub-block, then the test iris feature data of the g-th sub-block may belong to the template tester.

9. The low-computing-power scenario identity verification method based on iris information content and complex field features according to claim 8, characterized in that, In step S11, the method for determining whether a tester is a template tester is as follows: S111: If the real part of the tester and the real part of the template tester have the same representation in the g-th sub-block, then set the exponential logic judgment value H(g) of the g-th sub-block to 1. If the real part of the tester and the real part of the template tester are represented differently in the g-th sub-block, then the exponential logic judgment value H(g) of the g-th sub-block is set to 0. S112: Obtain the matching modulus distance LC between the tester and the template tester; to determine the template tester with the smallest matching modulus distance to the tester; In the formula: G represents the total number of sub-blocks under the optimal partitioning dimension; Let be the real part of the complex feature of the tester in the g-th sub-block; L represents the imaginary part of the complex feature of the tester in the g-th sub-block; g-1 Let be the real part of the complex feature of the template tester in the g-th sub-block; L g-2 Let be the imaginary part of the complex features of the template testers in the g-th sub-block; S113: If the template tester with the smallest matching modulus distance to the tester and the tester's exponential logic judgment value H(g) on ​​all sub-blocks are both equal to 1, and the matching modulus distance is less than the matching threshold μ, then the tester is determined to be the template tester with the smallest matching modulus distance to the tester.

Images