WO2017113083A1 - Method and apparatus for iris recognition - Google Patents

Abstract

A method and an apparatus for iris recognition for improvement of iris recognition rate. The method in the embodiments of the present invention comprises: obtaining an initial iris image of a subject; determining a Region of Interest (ROI) of the initial iris image; pre-processing the ROI to obtain a pre-processed iris image; processing the pre-processed iris image by using a 2D-Gabor filter to obtain a first set of feature data of the pre-processed iris image; processing the pre-processed iris image by using a Local Binary Pattern (LBP) algorithm to obtain a first set of feature vectors of the pre-processed iris image; calculating a first Hamming distance between the first set of feature data and prestored feature data, and calculating a first vector distance between the first set of feature vectors and prestored feature vectors; calculating weighted values of the first Hamming distance and the first vector distance, and recognizing ID of the subject according to the weighted values.

Description

Iris recognition method and device Technical field

The invention relates to a neighborhood of human biometrics technology, and particularly relates to an iris recognition method and device.

Background technique

The iris is part of the eye structure and is unique to the finger print and can be used to confirm identity. Although the iris itself has strong anti-counterfeiting features, the iris recognition technology itself is just a simple technique for photographing and character comparison. In the iris recognition technology, when the iris of the recognized person is photographed by the sensor, the recognized person may use a high-definition image, a fake iris image or even a 3D fake eyeball instead of the photographed photo to deceive the sensor, so that the criminals pass Iris recognition technology acquires important information from real users, resulting in loss of information or property.

In order to improve the safety of iris recognition technology, iris recognition technology is constantly improving to reduce the risk of illegal molecules forging iris.

Summary of the invention

The embodiment of the invention provides an iris recognition method and device for improving the accuracy of iris recognition, so as to solve the problem that the iris recognition technology in the prior art is not high in safety.

A first aspect of the present invention provides an iris recognition method, which may include:

Obtaining an initial iris image of the identified person;

Determining a region of interest ROI in the initial iris image;

Pre-processing the above ROI to obtain a pre-processed iris image;

Processing the pre-processed iris image by using a 2D-Gabor filter to obtain a first set of feature data of the pre-processed iris image;

The pre-processed iris image is processed by using a local binary mode LBP algorithm to obtain a first set of feature vectors of the preprocessed iris image;

Calculating a first Hamming distance of the first set of feature data and the pre-stored feature data, and calculating a first vector distance of the first set of feature vectors and the pre-stored feature vector, wherein the pre-stored feature data is previously determined according to the identified person The iris calculation calculates that the feature vector is calculated in advance according to the iris of the identified person;

Calculating a weighting value of the first Hamming distance and the first vector distance, according to the weighting value pair The identified person is identified.

It can be seen that in the embodiment of the present invention, the ROI is first determined from the initial iris image, and then the ROI is preprocessed to obtain a preprocessed iris image, and the feature data in the preprocessed iris image is obtained by using two different methods. The feature data of the pre-processed iris image is obtained by using 2D-Gabor filter. One is to obtain the feature vector of the pre-processed iris image by LBP algorithm, and then calculate the Hamming distance between the feature data and the pre-stored feature data, and calculate the feature vector and The vector distance of the pre-stored feature vector, the pre-stored feature data and the pre-stored feature vector are obtained from the real iris image of the recognized person, and the so-called real iris image is obtained directly from the eyes of the recognized person through the near-infrared sensor. Iris image, not a forged iris image obtained by high-definition printed images or 3D fake eyeballs. Then, the weighted value of Hamming distance and vector distance is calculated. The weighted value can accurately identify the identified person and the combination of the two layers of feature data, which can improve the recognition rate and reduce the risk of illegal molecules forging iris.

In some embodiments of the present invention, the obtaining the initial iris image of the recognized person comprises: acquiring an initial iris image of the recognized person by using a near-infrared sensor.

In some embodiments of the present invention, determining the ROI of the ROI in the initial iris image includes: determining a plurality of key points in the initial iris image, and determining an ROI in the initial iris image according to the plurality of key points.

Optionally, the above key points include six key points uniformly distributed on the boundary between the iris and the pupil, respectively being the first key point, the second key point, the third key point, the fourth key point, and the fifth key point. And the sixth key point; the above key points also include four key points distributed on the above-mentioned iris and eye white boundary, which are the 7th key point, the 8th key point, the 9th key point and the 10th Key point; wherein the first key point, the fourth key point and the ninth key point are on a straight line, the second key point, the fifth key point, the seventh key point and the tenth The key points are on a straight line, and the third key point, the sixth key point, and the eighth key point are on a straight line.

In some embodiments of the present invention, the pre-processing the ROI to obtain the pre-processed iris image comprises: performing polar coordinate transformation on the ROI to obtain a rectangular iris image; normalizing the rectangular iris image to obtain the pre-predetermined Handle the iris image.

In some embodiments of the present invention, the processing the first pre-processed iris image by using the 2D-Gabor filter to obtain the first set of feature data of the pre-processed iris image comprises: dividing the pre-processed iris image into M image regions; M is a positive integer greater than or equal to 2; using 2D-Gabor The filter convolves each of the M image regions to obtain a corresponding response amplitude; encodes the response amplitude, and combines the codes corresponding to the M image regions to obtain the first set of feature data.

Further, the above-mentioned image region of each of the M image regions is convoluted by using a 2D-Gabor filter to obtain a corresponding response amplitude, including: using K frequency L direction 2D-Gabor filters, for the above M Each image area in each image area is convoluted to obtain K×L response amplitude values corresponding to each image area;

At the same time, the above-mentioned response amplitude is encoded, and combining the codes corresponding to the M image regions to obtain the feature data includes: binarizing and encoding L response amplitudes at each frequency in each image region to obtain Each of the two regions corresponding to each frequency of the image region, combining K frequencies in each image region to obtain K×2 codes corresponding to each image region; combining the codes corresponding to the M image regions to obtain the first group The feature data, the first set of feature data includes M × K × 2 codes.

Further, the above-mentioned L response amplitude values at each frequency in each image region are binarized, and two codes corresponding to each frequency of each image region are obtained: when the above L response frames The nth response amplitude of the value is not greater than the n+1th response amplitude, and the nth response amplitude is correspondingly binarized to 1, when the nth response amplitude is greater than the n+1th In response to the amplitude, the nth response amplitude is correspondingly binarized to 0, wherein the n is a positive integer greater than or equal to 1; combining the codes corresponding to the L response amplitudes to obtain L binary values Encoding; obtaining two codes according to the above L binarization codes, and combining K codes of each of the image regions to obtain K×2 codes.

In some embodiments of the present invention, the processing of the pre-processed iris image by using the local binary mode LBP algorithm to obtain the first set of feature vectors of the pre-processed iris image comprises: dividing the pre-processed iris image into N image regions, The N is a positive integer greater than or equal to 2; the LBP feature corresponding to each of the N image regions is obtained, and the LBP features corresponding to the N image regions are combined to obtain the first set of feature vectors.

Further, the obtaining the LBP feature corresponding to each of the N image regions and combining the LBP features corresponding to the N image regions to obtain the first set of feature vectors includes: performing, for each pixel in each image region Binary coding to obtain the binary value corresponding to each pixel The coding method is obtained according to the binarization coding corresponding to all the pixels in each image region, and the histogram corresponding to each of the image regions is obtained, and the first group of feature vectors is obtained by combining the histograms corresponding to the N image regions.

Further, the above-mentioned binarization coding is performed on each pixel in each image region, and obtaining the binarization code corresponding to each pixel includes: acquiring gray values of each pixel in each image region, and sequentially comparing them. The gray value of each pixel and the gray value of 8 neighborhood pixels; when the gray value of the pixel is greater than the gray value of the neighborhood pixel, the corresponding binarization code is 1; when the pixel is When the gray value is less than or equal to the gray value of the eight neighboring pixel points, the corresponding binarized code is 0, and the binarized code of the 8-bit byte corresponding to each pixel point is obtained.

In some embodiments of the present invention, before calculating the vector distance between the first set of feature vectors and the pre-stored feature vectors, the method further comprises: performing dimensionality reduction on the feature vectors, and normalizing the reduced-dimensional feature vectors to obtain Normalizing the feature vector; further, calculating the vector distance between the feature vector and the pre-stored feature vector includes: calculating a vector distance between the normalized feature vector and the pre-stored feature sequence.

In some embodiments of the present invention, before calculating the weighting value of the first Hamming distance and the first vector distance, the method comprises: dividing the pre-processed iris image into H image regions, using one frequency and J directions. 2D-Gabor processes the H image regions to obtain a second set of data features; calculates a second Hamming distance between the second set of data features and the pre-stored feature data; and further, calculating the Hamming distance and the vector The weighted value of the distance includes: a weighting value for calculating the first Hamming distance, the second Hamming distance, and the first vector feature.

A second aspect of the present invention provides an iris recognition apparatus, which may include:

An obtaining module, configured to acquire an initial iris image of the identified person;

a determining module, configured to determine a region of interest ROI in the initial iris image;

a preprocessing module, configured to preprocess the ROI to obtain a preprocessed iris image;

a feature acquisition module, configured to process the pre-processed iris image by using a 2D-Gabor filter to obtain a first set of feature data of the pre-processed iris image; and processing the pre-processed iris image by using a local binary mode LBP algorithm to obtain the pre-processing a first set of eigenvectors of the iris image;

An identification module, configured to calculate a first Hamming distance of the first set of feature data and the pre-stored feature data, and calculate a first vector distance between the first set of feature vectors and the pre-stored feature vector, where the pre-stored The stored feature data is calculated in advance according to the iris of the identified person, and the feature vector is calculated according to the iris of the recognized person in advance; calculating a weighting value of the distance between the first Hamming distance and the first vector, according to the weighting The value identifies the identified person.

In some embodiments of the present invention, the acquiring module is specifically configured to acquire an initial iris image of the identified person by using a near-infrared sensor.

In some embodiments of the present invention, the determining module is specifically configured to determine a plurality of key points in the initial iris image, and determine an ROI in the initial iris image according to the plurality of key points.

Optionally, the above key points include six key points uniformly distributed on the boundary between the iris and the pupil, respectively being the first key point, the second key point, the third key point, the fourth key point, and the fifth key point. And the sixth key point; the above key points also include four key points distributed on the above-mentioned iris and eye white boundary, which are the 7th key point, the 8th key point, the 9th key point and the 10th Key point; wherein the first key point, the fourth key point and the ninth key point are on a straight line, the second key point, the fifth key point, the seventh key point and the tenth The key points are on a straight line, and the third key point, the sixth key point, and the eighth key point are on a straight line.

In some embodiments of the present invention, the pre-processing module is specifically configured to perform polar coordinate transformation on the ROI to obtain a rectangular iris image; and normalize the rectangular iris image to obtain the pre-processed iris image.

In some embodiments of the present invention, the feature acquiring module is further configured to: divide the pre-processed iris image into M image regions; the M is a positive integer greater than or equal to 2; and use the 2D-Gabor filter to Each of the image regions is convoluted to obtain a corresponding response amplitude; the response amplitude is encoded, and the first group of feature data is obtained by combining the codes corresponding to the M image regions.

In some embodiments of the present invention, the feature acquiring module is further specifically configured to perform convolution on each of the M image regions by using K frequency L direction 2D-Gabor filters to obtain each of the image regions. K×L response amplitudes corresponding to image regions; binarizing and encoding L response amplitudes at each frequency in each image region to obtain two codes corresponding to each frequency of each image region, Combining K frequencies in each image region to obtain K×2 codes corresponding to each image region; combining the corresponding images of the M image regions to obtain the first group of feature data, wherein the first group of feature data includes M×K× 2 codes.

In some embodiments of the present invention, the feature acquiring module is further configured to: when the nth response amplitude of the L response amplitudes is not greater than the n+1th response amplitude, the nth response amplitude Corresponding to the binarization code is 1, when the nth response amplitude is greater than the n+1th response amplitude, the nth response amplitude is correspondingly binarized to 0, wherein the n is greater than Or a positive integer equal to 1; combining the codes corresponding to the L response amplitudes to obtain L binarized codes; obtaining 2 codes according to the L binarized codes, combining K frequencies in each image region The encoding yields K x 2 codes.

In some embodiments of the present invention, the feature acquiring module is further configured to divide the pre-processed iris image into N image regions, where the N is a positive integer greater than or equal to 2; and each of the N image regions is acquired. The LBP feature corresponding to the image region combines the LBP features corresponding to the N image regions to obtain the first set of feature vectors.

In some embodiments of the present invention, the feature acquiring module is further configured to perform binarization coding on each pixel in each image region to obtain a binarized code corresponding to each pixel; according to all the image regions. Binary coding corresponding to the pixel, obtaining a histogram corresponding to each image region; combining the histogram corresponding to the N image regions to obtain the first group of feature vectors.

In some embodiments of the present invention, the feature acquiring module is further configured to: acquire gray values of each pixel in each image region, and sequentially compare gray values of each pixel with 8 neighbor pixels. Gray value; when the gray value of the pixel is greater than the gray value of the neighborhood pixel, the corresponding binarization code is 1; when the gray value of the pixel is less than or equal to the gray of 8 neighborhood pixels For the value, the corresponding binarization code is 0, and the binarized code of the 8-bit byte corresponding to each pixel is obtained.

In some embodiments of the present invention, the foregoing identification module is specifically configured to perform dimension reduction processing on the feature vector, and perform normalization processing on the reduced dimension feature vector to obtain a normalized feature vector; and calculate the normalization. The vector distance of the feature vector from the pre-stored feature sequence described above.

In some embodiments of the present invention, the feature acquiring module is further configured to divide the pre-processed iris image into H image regions, and process the H image regions by using 2D-Gabors with 1 frequency and J directions. a second set of data features; further, the identifying module is configured to calculate a second Hamming distance between the second set of data features and the pre-stored feature data; and calculate the first Hamming distance, the second Hamming distance, and The weighting value of the first vector feature described above.

A third aspect of the present invention provides an iris recognition apparatus, which may include:

Processor and memory;

The memory is used to store a program;

The processor is configured to execute a program in the memory such that the iris recognition device performs the iris recognition method provided by the first aspect.

A fourth aspect of the invention provides a storage medium storing one or more programs, the one or more programs comprising instructions that are executed by an iris recognition device of a third aspect comprising one or more processors The iris recognition device is caused to perform the iris recognition method provided by the first aspect.

DRAWINGS

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following description of the embodiments will be briefly described. It is obvious that the drawings in the following description are only some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without paying any creative work.

1 is a schematic flowchart of an iris recognition method according to an embodiment of the present invention;

2a is a schematic diagram of key points provided by an embodiment of the present invention;

2b is a schematic diagram of an ROI area according to an embodiment of the present invention;

2c is a schematic diagram of a rectangular ROI according to an embodiment of the present invention;

2d is a schematic diagram of a preprocessed iris image according to an embodiment of the present invention;

2e is an effect diagram of a Gabor filter kernel function in different parameter configurations according to an embodiment of the present invention;

2f is a diagram of relationship between a pixel point and eight neighboring pixel points according to an embodiment of the present invention;

3 is a schematic structural diagram of an iris recognition device according to an embodiment of the present invention;

FIG. 4 is another schematic structural diagram of an iris recognition device according to an embodiment of the present invention.

detailed description

The technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the accompanying drawings in the embodiments of the present invention. It is obvious that the described embodiments are only a part of the embodiments of the present invention, but not all embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

Embodiments of the present invention provide an iris recognition method for improving iris recognition rate and reducing the risk of illegal molecules forging iris. The embodiment of the invention further provides an apparatus corresponding to the iris recognition method.

The iris is part of the eye structure, and the center of the iris has a circular opening called a through hole. Because everyone The irises are all different, so the iris can be used for identification. The iris is different in color, and its color is different. It is mainly blue and brown, and other colors are mixed. Among them, the white iris is mostly blue, while the yellow iris is mostly brown. The blue iris is clear under visible light, but the brown iris is not clear under visible light, while the brown iris can be under near infrared. clear and distinct. Therefore, for the yellow iris of the yellow race, in the embodiment of the present invention, the near-infrared sensor is used to obtain a clear iris image.

When identifying an identity through the iris, it is mainly by analyzing the iris image. The iris image can be taken directly from the eyes of the identified person, as well as illegal molecules to fake the identified image by forging the iris image. The forged iris image can be an iris image printed by high-definition technology or an iris image obtained by photographing a 3D fake eyeball. The forged iris image has a great similarity to the real iris image, and the iris recognition system can easily treat these forged iris images as real iris images.

However, due to the high-resolution color iris image printed by using an inkjet printer, since the absorption of the near-infrared spectrum is the same for the printing inks of various colors, it is difficult to photograph the eyes in the near-infrared light-sensing camera. Iris image. The 3D fake eyes made of silica gel or other materials have a great difference between the materials and the real eyes, which are embodied in the nuances of some features. Therefore, the processing of the feature data can be enhanced when the iris recognition is performed to improve The accuracy of iris recognition reduces the risk of being forged.

Based on the above discussion, please refer to FIG. 1. FIG. 1 is a schematic flowchart diagram of an iris recognition method according to an embodiment of the present invention; as shown in FIG. 1, an iris recognition method may include:

101. Acquire an initial iris image of the identified person;

The embodiment of the present invention can be preferably used to identify the identity of a yellow person. Therefore, it is important to target a brown iris. In order to obtain a clearer iris image, in the embodiment of the present invention, the device for collecting the iris image may be a near-infrared sensor. Photographing through a near-infrared sensor to obtain an initial iris image of the identified person.

102. Determine a Region of Interest (ROI) in the initial iris image.

Firstly, the ROI in the initial iris image is determined, and the identity is determined according to the features in the ROI, which can improve the accuracy of the recognition.

In some embodiments of the invention, a number of key points may be determined in the initial iris image first. These key points can be on the pupil's pupil-iris boundary and on the dividing line between the iris and the white of the eye. Referring to FIG. 2a, FIG. 2a is a schematic diagram of key points provided by an embodiment of the present invention. In FIG. 2a, 10 key points are determined in the initial iris image, wherein 6 key points are obtained at the boundary between the pupil and the iris. These six key points are respectively distributed at the boundary between the pupil and the iris, which are the first key point, the second key point, the third key point, the fourth key point, the fifth key point and the sixth key point. The other four key points are at the boundary between the iris and the white of the eye, which are 7 key points, 8th key point, 9th key point and 10th key point. As can be seen in Figure 2a, the first key point, the fourth key point and the ninth key point are on a straight line, the second key point, the fifth key point, and the seventh key. The point and the 10th key point are on a straight line, and the 3rd key point, the 6th key point, and the 8th key point are in a straight line.

According to the above 10 key points, an ROI area is determined. As shown in FIG. 2b, the determined ROI is the iris area between the pupil and the white of the eye. As shown in the white area in FIG. 2b, it can be seen that The ROI determined from the initial iris image is an annular region.

103. Perform pre-processing on the ROI to obtain a pre-processed iris image;

The preprocessing of the ROI may include a polar coordinate transformation of the ROI. It can be understood that the iris surrounding the pupil is also an annular region. In the preprocessing, the circular ROI needs to be transformed into a rectangular ROI. Specifically, the annular ROI may be cut at a diameter and then nonlinearly stretched such that the circular ROI transformation is referred to as a rectangular ROI, and as shown in Figure 2c, the ROI is stretched to obtain a rectangular region.

Further, normalizing the rectangular ROI, the rectangular ROI is transformed into a fixed-size rectangular ROI, and the localized grayscale contrast is normalized to the normalized rectangular ROI, as shown in FIG. 2d. Finally, a preprocessed iris image is obtained.

104. The preprocessed iris image is processed by using a 2D-Gabor filter to obtain a first set of feature data of the preprocessed iris image;

It should be noted that the Gabor filter can extract the image frequency and directional characteristics sensitive to the human visual system. Therefore, in the embodiment of the present invention, the Gabor filter can be used to filter the ROI to obtain better results. .

In the embodiment of the present invention, by sampling a plurality of iris images, the iris image is divided into a plurality of iris sub-modules, and then the iris sub-module is processed to obtain a 2D Gabor filter.

The calculation formula of the 2D Gabor filter is as follows:

u=x cosθ+y sinθ

v=y cosθ-x sinθ

Where θ is the direction of the filter, δ _u is the standard deviation of the Gaussian envelope parallel to the θ direction, and δ _v is the standard deviation of Gaussian included in the direction perpendicular to θ, where the value can be 1, ω The frequency of the complex sine function. Referring to FIG. 2e, FIG. 2e is an effect diagram of a Gabor filter kernel function in different parameter configurations according to an embodiment of the present invention.

As shown in Figure 2e, Figure (a) is an image of the Gabor filter kernel function at θ = 0; Figure (b) is

Image of the Gabor filter kernel function; Figure (c) is

Image of the Gabor filter kernel function; Figure (d) is

The image of the Gabor filter kernel function; Fig. (e) is the image of the Gabor filter kernel function when θ = π; Fig. (f) is the image of the Gabor filter kernel function when π = 0.1; Fig. (g) is π = The image of the Gabor filter kernel function at 0.3°; and (h) is the image of the Gabor filter kernel function at θ=0.

The study found that Gabor filters are well suited for texture expression and separation. In the spatial domain, a 2DGabor filter is a Gaussian kernel function modulated by a sinusoidal plane wave. After the ROI is filtered, the feature data corresponding to the ROI may be obtained. One specific manner of filtering the ROI may be obtaining the feature data by a convolution operation.

In the embodiment of the present invention, the response amplitude is first obtained by a 2D Gabor filter, and then the response amplitude is encoded to obtain feature data.

After many experiments, when the frequency of the complex sinusoid function of a set of Gabor filters is ω ₁ =4.0, and the frequency of the complex sinusoidal function of another set of Gabor filters is ω ₂ =4.5, better results can be obtained. More recognizable. Wherein, the image area here may be an image area in which two adjacent image areas have partial overlap.

For example, according to the above description of the 2D Gabor filter, each region image is processed using a 2D Gabor filter with 6 frequencies and 16 directions. First, you can first calculate two sets of data from two frequencies, the Gabor filter of 16 directions for each frequency. Assuming that two frequencies are used, the Gabor filter of each direction is filtered, and the effective part is a convolution window of 10×6 size. Then the data is divided into two groups according to the frequency, which are:

The preprocessed image is filtered. From the upper left corner, an image block of the same size as the convolution window is convolved with each set of convolution windows, and the convolution window size is 10×6, thereby obtaining two sets of 16 response amplitude values.

Binary coding is performed on the response amplitude value to obtain binarized coding. Binary encoding the response amplitude value to obtain a 16-bit 0-1 code

Where 1 () indicates that the expression in parentheses is "true", the value is 1; otherwise, the value is 0. Integer in the above formula

In fact, it is less than 216. Therefore, it can be expressed as 2 bytes of feature data, in which case an image block is represented as 4 bytes of feature data.

After that, the 16-bit 0-1 binarization coding is composed of two bytes of data in a binary arrangement, that is, two binarization codes, and then 2×2 binarization codes can be obtained at two frequencies, that is, 4 The feature data of the bytes, if M is 10×6 overlapping or overlapping image areas, then 60×4=240 bytes of feature data, that is, the first set of feature data, will be obtained.

105. The pre-processed iris image is processed by using a Local Binary Patterns (LBP) algorithm to obtain a first set of feature vectors of the pre-processed iris image;

It can be understood that LBP is an effective texture description operator, which measures and extracts the local texture information of the image and has invariance to the illumination. The Unified LBP occupies the vast majority of all modes in the image. The different sampling radii and the number of surrounding pixels will be different. Therefore, Unified LBP has a better description effect on the local texture description.

Based on the advantages of Unified LBP, the pre-processed iris image is processed using Unified LBP in the embodiment of the present invention to obtain a feature vector.

For example, the pre-processed iris image is divided into 7×3 image regions, which may be image regions having no overlapping portions between adjacent two regions.

In each image region, as shown in Fig. 2f, for each pixel (as 0 in Fig. 2f), it is in the middle of its eight neighborhood pixels (1 to 8 in Fig. 2f, respectively). Compare the pixel points in turn The gray value and the gray value of the eight neighboring pixel points (1 to 8) can be compared in time or in a counterclockwise manner, when the gray value of the pixel is higher than the gray value of the neighboring pixel When large, the corresponding binary encoding is 1; conversely, when the gray value of the pixel is less than or equal to the gray value of the neighboring pixel, the corresponding binarized encoding is 0, and so on, the pixel is obtained. Corresponding binabyte encoding of 1 octet. For example, in FIG. 2f, the comparison is performed in the order of the pointer, the gray value of the pixel point 0 is larger than the gray value of the pixel point 1, and the corresponding code is one; the gray value of the pixel point 0 is larger than the gray value of the pixel point 2, corresponding to Encoding a 1; the gray value of pixel 0 is not greater than the gray value of pixel 3, corresponding to a 0; the gray value of pixel 0 is not greater than the gray value of pixel 4, corresponding to a 0; pixel The gray value of 0 is not greater than the gray value of pixel 5, corresponding to a code of 0; the gray value of pixel 0 is greater than the gray value of pixel 6, corresponding to a code of 1; the gray value of pixel 0 is not greater than The gray value of pixel 7 corresponds to a code of 0; the gray value of pixel 0 is greater than the gray value of pixel 8, and the corresponding code encodes a 1, then the pixel will get a corresponding 8-byte binary code. For: 11000101.

After binarizing and encoding all the pixels in each image region, a histogram is obtained according to the binarization coding of all the pixel points, and the histograms of all image regions are concatenated to obtain a feature vector.

According to the above division into 7×3 image regions, a feature vector of 21×58=1218 bytes can be obtained, and then the feature vector of 1218 bytes is subjected to dimensionality reduction processing to obtain a 20-dimensional feature vector, which is then normalized. It is between 0 and 255.

106. Calculate a first Hamming distance of the first set of feature data and pre-stored feature data, and calculate a first vector distance between the first set of feature vectors and a pre-stored feature vector, where the pre-stored feature data is based on Calculated by the iris of the identified person, the feature vector is calculated in advance according to the iris of the identified person;

Of course, the feature data of the identified person is stored in the database, and when the identity of the identified person is recognized, the Hamming distance between the feature data and the pre-stored feature data is calculated, as follows:

The similarity between the feature data and the pre-stored feature data is calculated by the following formula:

Where x ₁ is the above characteristic data, x ₂ is pre-stored feature data stored in the database, and S(x ₁ , x ₂ ) is the similarity between the feature data and the pre-stored feature data, H(x ₁ , x ₂ ) For the Hamming distance between x ₁ and x ₂ , F and B represent two constants, respectively.

In information coding, the number of bits encoded on the corresponding bits of two legal codes is called the code distance, also known as the Hamming distance. In a valid coding set, the minimum Hamming distance of any two codewords is called the Hamming distance of the code set. For example: 10101 and 00110 have the first, fourth, and fifth positions from the first place, and the Hamming distance is 3.

F and B respectively represent two constants, which can be confirmed by experiments. The specific measurement method can be to calculate the values of F and B by simulation.

In addition, the feature vector of the identified person is also stored in the database. Then, the similarity can be judged by calculating the vector distance between the above feature vector and the pre-stored feature vector.

107. Calculate a weighting value of the first Hamming distance and the first vector distance, and identify the identified person according to the weighting value.

In order to improve the recognition efficiency, and the focus of acquiring feature data by 2D Gabor is different from that of using Unified LBP to acquire feature vectors, the feature data acquired by 2D Gabor is different from the feature vector obtained by using Unified LBP. . In the embodiment of the present invention, the feature data and the feature vector acquired in the two modes are further weighted to obtain a weight value, and then the weighted value is used to identify the identity of the identified person.

For example, the Hamming distance is 50 and the vector distance is 50. The weighting value is calculated as follows:

50×2/2=50

It can be seen that in the embodiment of the present invention, the ROI is first determined from the initial iris image, and then the ROI is preprocessed to obtain a preprocessed iris image, and the feature data in the preprocessed iris image is obtained by using two different methods. The first set of feature data is obtained by using a 2D-Gabor filter to obtain a preprocessed iris image. One is to obtain a first set of feature vectors of the preprocessed iris image by using the LBP algorithm, and then calculate the first set of feature data and pre-stored feature data. The Hamming distance is calculated by calculating the vector distance between the first set of feature vectors and the pre-stored feature vectors. The pre-stored feature data and the pre-stored feature vectors are obtained from the real iris image of the recognized person. The so-called real iris image refers to passing near The infrared sensor directly captures the iris image from the eyes of the identified person, rather than the forged iris image obtained by high-definition printed images or 3D fake eyeballs. Then, the weighted value of Hamming distance and vector distance is calculated. The weighted value can accurately identify the identified person and the combination of the two layers of feature data, which can improve the recognition rate and reduce the risk of illegal molecules forging iris.

In addition, in the foregoing embodiment, the identity of the identified person is identified by the weighted value of the Hamming distance and the vector distance, and the feature of the initial iris image is more correctly reflected for the weighted value. In the embodiment of the present invention, the identifier may further be further The preprocessed iris image is divided into 13×5 image regions without overlapping portions according to 13×5, and then each image region of 13×5 image regions is processed by 2D Gabor, and the process can refer to the above steps. The specific description in 104 results in a second set of feature data of 65 x 4 = 260 bytes.

Then, in step 106, it is also required to calculate a Hamming distance between the second set of feature data of 65×4=260 bytes and the pre-stored feature data, where the Hamming distance is referred to as the Hamming distance B, and the above embodiment The Hamming distance obtained in the middle is called Hamming distance A. Further, in the step, the weighted values of the Hamming distance A, the Hamming distance B, and the vector distance are calculated. Since a weight is added, the calculated weighting value can more accurately reflect the characteristics of the initial iris image, thereby improving the recognition accuracy.

Further, in the embodiment of the present invention, the pre-processed iris image may be divided into different numbers of image regions multiple times, and the multiple division manners may be partially overlapped or overlapped, and then the above 2D Gabors are respectively utilized. Processing is performed to obtain a plurality of sets of feature data, thereby obtaining a corresponding number of Hamming distances. Similarly, the preprocessed iris image is again divided into different numbers of image regions, and then the divided image regions are processed by Unified LBP to obtain a plurality of feature vectors, thereby obtaining a corresponding number of vector distances. Finally, multiple Hamming distances and multiple vector distances are weighted to obtain a weighted value.

It can be seen that in weighting, the more weights are obtained, the more accurate the weighting value is, that is, the pre-processed iris image needs to be processed multiple times using 2D Gabor or Unified LBP.

Referring to FIG. 3, FIG. 3 is a schematic structural diagram of an iris recognition device according to an embodiment of the present invention; as shown in FIG. 3, an iris recognition device may include:

An obtaining module 310, configured to acquire an initial iris image of the identified person;

a determining module 320, configured to determine a region of interest ROI in the initial iris image;

a preprocessing module 330, configured to preprocess the ROI to obtain a preprocessed iris image;

a feature acquiring module 340, configured to process the preprocessed iris image by using a 2D-Gabor filter to obtain a first set of feature data of the preprocessed iris image; and process the preprocessed iris image by using a local binary mode LBP algorithm, Obtaining a first set of feature vectors of the preprocessed iris image;

The identification module 350 is configured to calculate the first set of feature data and the first sea of pre-stored feature data a clear distance, and calculating a first vector distance between the first set of feature vectors and a pre-stored feature vector, the pre-stored feature data being calculated in advance according to an iris of the recognized person, the feature vector being previously according to the Calculating the iris of the identified person; calculating a weighting value of the distance between the first Hamming distance and the first vector, and identifying the identified person according to the weighting value.

It can be seen that, in the embodiment of the present invention, the initial iris image of the recognized person is acquired by the obtaining module 310, and then the determining module 320 obtains the ROI from the initial iris image acquired by the obtaining module 310, and the pre-processing module 330 firstly compares the ROI. Performing pre-processing to obtain a pre-processed iris image, and then acquiring the first set of feature data by the feature acquisition module 340 using the 2D-Gabor filter, and acquiring the first set of feature vectors by using the LBP algorithm, and the recognition module 350 calculates the first set by The Hamming distance of the feature data and the pre-stored feature data and the vector distance between the first set of feature vectors and the pre-stored feature vector, and further weighting the Hamming distance and the vector distance to obtain a weight value, and identifying the identity of the identified person by the weighted value .

In some embodiments of the present invention, the acquiring module 310 is specifically configured to acquire an initial iris image of the identified person by using a near-infrared sensor.

In some embodiments of the present invention, the determining module 320 is specifically configured to determine a plurality of key points in the initial iris image, and determine an ROI in the initial iris image according to the plurality of key points.

Optionally, the above key points include six key points uniformly distributed on the boundary between the iris and the pupil, respectively being the first key point, the second key point, the third key point, the fourth key point, and the fifth key point. And the sixth key point; the above key points also include four key points distributed on the above-mentioned iris and eye white boundary, which are the 7th key point, the 8th key point, the 9th key point and the 10th Key point; wherein the first key point, the fourth key point and the ninth key point are on a straight line, the second key point, the fifth key point, the seventh key point and the tenth The key points are on a straight line, and the third key point, the sixth key point, and the eighth key point are on a straight line.

In some embodiments of the present invention, the pre-processing module 330 is specifically configured to perform polar coordinate transformation on the ROI to obtain a rectangular iris image; normalize the rectangular iris image to obtain the pre-processed iris image.

In some embodiments of the present invention, the feature acquiring module 340 is further specifically configured to divide the preprocessed iris image into M image regions; the M is a positive integer greater than or equal to 2; and the 2D-Gabor filter is used to Each of the M image regions is convoluted to obtain a corresponding response amplitude; the response amplitude is encoded to combine the M image regions corresponding to The first set of feature data is obtained by encoding.

In some embodiments of the present invention, the feature acquiring module 340 is further specifically configured to perform convolution on each of the M image regions by using K frequency L direction 2D-Gabor filters. K × L response amplitudes corresponding to one image region; binarizing the L response amplitudes at each frequency in each image region to obtain two codes corresponding to each frequency of each image region Combining K frequencies in each image region to obtain K×2 codes corresponding to each image region; combining corresponding M image regions to obtain the first group of feature data, wherein the first group of feature data includes M×K × 2 codes.

In some embodiments of the present invention, the feature acquiring module 340 is further configured to: when the nth response amplitude of the L response amplitudes is not greater than the n+1th response amplitude, the nth response amplitude The value corresponding to the binarization code is 1, and when the nth response amplitude is greater than the n+1th response amplitude, the nth response amplitude is correspondingly binarized to 0, wherein the n is a positive integer greater than or equal to 1; combining the codes corresponding to the L response magnitudes to obtain L binarized codes; obtaining 2 codes according to the L binarized codes, combining K frequencies in each image region The encoding gives K x 2 encodings.

In some embodiments of the present invention, the feature acquiring module 340 is further configured to divide the pre-processed iris image into N image regions, where the N is a positive integer greater than or equal to 2; and each of the N image regions is acquired. The LBP feature corresponding to one image region combines the LBP features corresponding to the N image regions to obtain the first set of feature vectors.

In some embodiments of the present invention, the feature acquiring module 340 is further specifically configured to perform binarization coding on each pixel in each image region to obtain a binarized code corresponding to each pixel; according to each image region. Binary coding corresponding to all pixels, obtaining a histogram corresponding to each image region; combining the histograms corresponding to the N image regions to obtain the first group of feature vectors.

In some embodiments of the present invention, the feature acquiring module 340 is further configured to acquire gray values of each pixel in each image region, and sequentially compare gray values of each pixel with eight neighbor pixels. The gray value of the point; when the gray value of the pixel is greater than the gray value of the neighborhood pixel, the corresponding binarization code is 1; when the gray value of the pixel is less than or equal to the gray of the 8 neighborhood pixels In the case of the degree value, the corresponding binarization code is 0, and the binarized code of the 8-bit byte corresponding to each pixel point is obtained.

In some embodiments of the present invention, the foregoing identification module 350 is specifically configured to perform dimensionality reduction processing on the feature vector, and normalize the reduced-dimensional feature vector to obtain a normalized feature vector; The vector distance between the feature vector and the pre-stored feature sequence.

In some embodiments of the present invention, the feature acquiring module 340 is further configured to divide the pre-processed iris image into H image regions, and process the H image regions by using 2D-Gabors of one frequency and J directions. Obtaining a second set of data features; further, the identifying module 350 is configured to calculate a second Hamming distance between the second set of data features and the pre-stored feature data; and calculate the first Hamming distance and the second Hamming The distance and the weighting value of the first vector feature described above.

4 is another schematic structural diagram of an iris recognition apparatus according to an embodiment of the present invention, which may include at least one processor 401 (for example, a CPU, Central Processing Unit), at least one network interface or other communication interface, a memory 402, and at least one A communication bus for implementing connection communication between these devices. The processor 401 is configured to execute an executable module, such as a computer program, stored in a memory. The memory 402 may include a high speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system gateway and at least one other network element is implemented by at least one network interface (which may be wired or wireless), and an Internet, a wide area network, a local network, a metropolitan area network, etc. may be used.

As shown in FIG. 4, in some embodiments, program instructions are stored in the memory 402, and the program instructions may be executed by the processor 401. The processor 401 specifically performs the following steps: acquiring an initial iris image of the identified person; Determining a region of interest ROI in the initial iris image; preprocessing the ROI to obtain a preprocessed iris image; processing the preprocessed iris image with a 2D-Gabor filter to obtain a preprocessed iris image a set of feature data; processing the preprocessed iris image by using a local binary mode LBP algorithm to obtain a first set of feature vectors of the preprocessed iris image; and calculating a first set of feature data and pre-stored feature data a Hamming distance, and calculating a first vector distance between the first set of feature vectors and a pre-stored feature vector, the pre-stored feature data being calculated in advance according to an iris of the recognized person, the feature vector being a prior basis Calculating the iris of the identified person; calculating a weighting value of the distance between the first Hamming distance and the first vector, according to the weighting value The identified person is identified.

In some embodiments, the processor 401 can also perform the following steps: passing near infrared rays a sensor that acquires an initial iris image of the identified person.

In some embodiments, the processor 401 can also perform the steps of determining a number of key points in the initial iris image, and determining an ROI in the initial iris image based on the plurality of key points.

In some embodiments, the processor 401 may further perform the following steps: performing polar coordinate transformation on the ROI to obtain a rectangular iris image; normalizing the rectangular iris image to obtain the preprocessed iris image .

In some embodiments, the processor 401 may further perform the steps of: dividing the pre-processed iris image into M image regions; the M is a positive integer greater than or equal to 2; using a 2D-Gabor filter pair Each of the M image regions is convoluted to obtain a corresponding response amplitude; the response amplitude is encoded, and the code corresponding to the M image regions is combined to obtain the first set of feature data. .

In some embodiments, the processor 401 may further perform the following steps: convolving each of the M image regions by using K frequency L direction 2D-Gabor filters to obtain each Corresponding K×L response amplitudes of the image region; the encoding the amplitude of the response, and combining the encoding corresponding to the M image regions to obtain the feature data comprises: for each frequency in each image region The lower L response amplitudes are binarized to obtain two codes corresponding to each frequency of each image region, and K frequencies in each image region are combined to obtain K×2 codes corresponding to each image region. Combining the codes corresponding to the M image regions to obtain the first set of feature data, the first set of feature data including M×K×2 codes.

In some embodiments, the processor 401 may further perform the step of: when the nth response amplitude of the L response amplitudes is not greater than the n+1th response amplitude, the nth response The amplitude corresponding to the binarization code is 1, and when the nth response amplitude is greater than the n+1th response amplitude, the nth response amplitude is correspondingly binarized to 0, wherein And n is a positive integer greater than or equal to 1; combining the codes corresponding to the L response amplitudes to obtain L binarized codes; and obtaining 2 codes according to the L binarized codes, combining each The encoding of the K frequencies in the image region yields K x 2 codes.

In some embodiments, the processor 401 may further perform the step of dividing the pre-processed iris image into N image regions, the N being a positive integer greater than or equal to 2; The LBP features corresponding to each of the N image regions are combined with the LBP features corresponding to the N image regions to obtain the first set of feature vectors.

In some embodiments, the processor 401 may further perform the following steps: performing binarization coding on each pixel in each image region to obtain a binarization code corresponding to each pixel; according to all the image regions Binary coding corresponding to the pixel, obtaining a histogram corresponding to each image region; combining the histogram corresponding to the N image regions to obtain the first group of feature vectors.

In some embodiments, the processor 401 may further perform the steps of: acquiring gray values of each pixel in each image region, and sequentially comparing gray values of each pixel with 8 neighborhood pixels. Gray value; when the gray value of the pixel is greater than the gray value of the neighborhood pixel, the corresponding binarization code is 1; when the gray value of the pixel is less than or equal to the gray of 8 neighborhood pixels For the value, the corresponding binarization code is 0, and the binarized code of the 8-bit byte corresponding to each pixel is obtained.

In some embodiments, the processor 401 may further perform the following steps: performing dimension reduction processing on the feature vector, and normalizing the reduced-dimensional feature vector to obtain a normalized feature vector.

In some embodiments, the processor 401 can also perform the step of calculating a vector distance of the normalized feature vector from the pre-stored feature sequence.

In some embodiments, the processor 401 may further perform the steps of: dividing the pre-processed iris image into H image regions, and performing the H image regions by using 2D-Gabors of 1 frequency and J directions; Processing, obtaining a second set of data features; calculating a second Hamming distance of the second set of data features and the pre-stored feature data; the calculating a weighted value of the Hamming distance and the vector distance comprises: calculating a weighted value of the first Hamming distance, the second Hamming distance, and the first vector feature.

Optionally, in some embodiments, the plurality of key points include six key points uniformly distributed on the boundary between the iris and the pupil, respectively being the first key point, the second key point, the third key point, and the fourth point The key points, the fifth key point and the sixth key point; the key points further include four key points distributed on the boundary between the iris and the white of the eye, respectively being the seventh key point, the eighth key point, The ninth key point and the tenth key point; wherein the first key point, the fourth key point, and the ninth key point are in a straight line, the second key point, the fifth key point The point, the 7th key point and the 10th key point are on a straight line, the 3rd key point, the 6th key point and the 8th key Point in a straight line.

In the above embodiments, the descriptions of the various embodiments are different, and the details that are not detailed in a certain embodiment can be referred to the related descriptions of other embodiments.

A person skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working process of the system, the device and the unit described above can refer to the corresponding process in the foregoing method embodiment, and details are not described herein again.

In the several embodiments provided by the present application, it should be understood that the disclosed system, apparatus, and method may be implemented in other manners. For example, the device embodiments described above are merely illustrative. For example, the division of the unit is only a logical function division. In actual implementation, there may be another division manner, for example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, and may be in an electrical, mechanical or other form.

The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a standalone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention, which is essential or contributes to the prior art, or all or part of the technical solution, may be embodied in the form of a software product stored in a storage medium. A number of instructions are included to cause a computer device (which may be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk. The medium in which the program code is stored.

The method and device for identifying an iris provided by the present invention are described in detail above. For the general technical personnel in the neighborhood, according to the idea of the embodiment of the present invention, there are changes in the specific implementation manner and application scope. In summary, the content of the specification should not be construed as limiting the invention.

Claims (28)

1. An iris recognition method, comprising:

   Obtaining an initial iris image of the identified person;

   Determining a region of interest ROI in the initial iris image;

   Pre-processing the ROI to obtain a pre-processed iris image;

   Processing the pre-processed iris image with a 2D-Gabor filter to obtain a first set of feature data of the pre-processed iris image;

   Processing the preprocessed iris image by using a local binary mode LBP algorithm to obtain a first set of feature vectors of the preprocessed iris image;

   Calculating a first Hamming distance of the first set of feature data and pre-stored feature data, and calculating a first vector distance of the first set of feature vectors and a pre-stored feature vector, wherein the pre-stored feature data is according to the foregoing Calculated by the iris of the identified person, the feature vector is calculated in advance according to the iris of the identified person;

   Calculating a weighting value of the first Hamming distance and the first vector distance, and identifying the identified person according to the weighting value.
2. The method of claim 1 wherein said obtaining an initial iris image of the identified person comprises:

   The initial iris image of the identified person is acquired by a near-infrared sensor.
3. The method according to claim 1 or 2, wherein the determining the region of interest ROI in the initial iris image comprises:

   A plurality of key points are determined in the initial iris image, and an ROI in the initial iris image is determined based on the plurality of key points.
4. The method of claim 3 wherein:

   The key points include six key points uniformly distributed on the boundary between the iris and the pupil, which are the first key point, the second key point, the third key point, the fourth key point, the fifth key point, and the sixth point. key point;

   The key points further include four key points distributed on the boundary between the iris and the white of the eye, which are the seventh key point, the eighth key point, the ninth key point and the tenth key point;

   Wherein the first key point, the fourth key point and the ninth key point are on a straight line, The second key point, the fifth key point, the seventh key point, and the tenth key point are on a straight line, and the third key point, the sixth key point, and the eighth key point are On a straight line.
5. The method according to any one of claims 1 to 4, wherein the preprocessing the ROI to obtain a preprocessed iris image comprises:

   Performing a polar coordinate transformation on the ROI to obtain a rectangular iris image;

   The rectangular iris image is normalized to obtain the preprocessed iris image.
6. The method according to claim 1, wherein the processing the pre-processed iris image by using a 2D-Gabor filter to obtain the first set of feature data of the pre-processed iris image comprises:

   Dividing the preprocessed iris image into M image regions; the M is a positive integer greater than or equal to 2;

   A 2D-Gabor filter is used to convolve each of the M image regions to obtain a corresponding response amplitude;

   The response amplitude is encoded, and the code corresponding to the M image regions is combined to obtain the first set of feature data.
7. The method according to claim 6, wherein the convolving each of the M image regions by using a 2D-Gabor filter to obtain a corresponding response amplitude comprises:

   Using K frequency L direction 2D-Gabor filters, convolving each of the M image regions to obtain K×L response amplitude values corresponding to each image region;

   The encoding the amplitude of the response, and combining the encoding corresponding to the M image regions to obtain the feature data includes:

   The L response amplitudes at each frequency in each image region are binarized to obtain two codes corresponding to each frequency of each image region, and the K frequencies in each image region are combined to obtain each K × 2 codes corresponding to one image region;

   The first set of feature data is obtained by combining codes corresponding to the M image regions, and the first set of feature data includes M×K×2 codes.
8. The method according to claim 7, wherein said L response amplitudes at each frequency in each image region are binarized to obtain two corresponding to each frequency of each image region. The codes include:

   When the nth response amplitude of the L response amplitudes is not greater than the n+1th response amplitude, the nth response amplitude is correspondingly binarized to 1 when the nth response When the amplitude is greater than the n+1th response amplitude, the nth response amplitude is correspondingly binarized to 0, where n is a positive integer greater than or equal to 1;

   Combining the codes corresponding to the L response amplitudes to obtain L binarization codes;

   Two codes are obtained according to the L binarization codes, and K×2 codes are obtained by combining the codes of K frequencies in each image region.
9. The method according to claim 1, wherein the processing the pre-processed iris image by using a local binary mode LBP algorithm, and obtaining the first set of feature vectors of the pre-processed iris image comprises:

   Dividing the preprocessed iris image into N image regions, wherein N is a positive integer greater than or equal to 2;

   And acquiring an LBP feature corresponding to each of the N image regions, and combining the LBP features corresponding to the N image regions to obtain the first group feature vector.
10. The method according to claim 9, wherein the acquiring an LBP feature corresponding to each of the N image regions, combining the LBP features corresponding to the N image regions, to obtain the first group Feature vectors include:

    Binarizing each pixel in each image region to obtain a binarized code corresponding to each pixel;

    Obtaining a histogram corresponding to each image region according to the binarization code corresponding to all the pixels in each image region;

    Combining the histograms corresponding to the N image regions, the first set of feature vectors is obtained.
11. The method according to claim 10, wherein the binarization coding is performed for each pixel in each image region, and the binarization code corresponding to each pixel is obtained:

    Obtaining a gray value of each pixel in each image region, and sequentially comparing the gray value of each pixel with the gray value of eight neighbor pixels;

    When the gray value of the pixel point is greater than the gray value of the neighboring pixel point, the corresponding binarization code is 1; when the gray value of the pixel point is less than or equal to the gray value of the 8 neighborhood pixels, the corresponding two The valued code is 0, and the binarized code of the 8-bit byte corresponding to each pixel is obtained.
12. The method according to claim 1 or 9 or 10 or 11, wherein the calculating the vector distance between the first set of feature vectors and the pre-stored feature vectors comprises:

    Performing dimensionality reduction processing on the feature vector, and normalizing the reduced dimension feature vector to obtain a normalized feature vector;

    The calculating a vector distance between the feature vector and the pre-stored feature vector includes:

    Calculating a vector distance of the normalized feature vector from the pre-stored feature sequence.
13. The method according to claim 1, wherein said calculating a weighting value of said first Hamming distance and said first vector distance comprises:

    Dividing the pre-processed iris image into H image regions, and processing the H image regions by using 2D-Gabors of 1 frequency and J directions to obtain a second set of data features;

    Calculating a second Hamming distance of the second set of data features and the pre-stored feature data;

    The calculating weighting values of the Hamming distance and the vector distance includes:

    A weighting value of the first Hamming distance, the second Hamming distance, and the first vector feature is calculated.
14. An iris recognition device, comprising:

    An obtaining module, configured to acquire an initial iris image of the identified person;

    a determining module, configured to determine a region of interest ROI in the initial iris image;

    a preprocessing module, configured to preprocess the ROI to obtain a preprocessed iris image;

    a feature acquiring module, configured to process the preprocessed iris image by using a 2D-Gabor filter to obtain a first set of feature data of the preprocessed iris image; and process the preprocessed iris image by using a local binary mode LBP algorithm to obtain a first set of feature vectors of the preprocessed iris image;

    An identification module, configured to calculate a first Hamming distance of the first set of feature data and pre-stored feature data, and calculate a first vector distance between the first set of feature vectors and a pre-stored feature vector, the pre-stored feature data Calculated in advance according to the iris of the identified person, the feature vector is calculated according to the iris of the identified person in advance; calculating a weighted value of the distance between the first Hamming distance and the first vector, according to The weighting value identifies the identified person.
15. The iris recognition device according to claim 14, wherein

    The acquiring module is specifically configured to acquire an initial iris image of the identified person by using a near-infrared sensor.
16. The iris recognition device according to claim 14 or 15, wherein

    The determining module is specifically configured to determine a plurality of key points in the initial iris image, and determine an ROI in the initial iris image according to the plurality of key points.
17. The iris recognition device according to claim 16, wherein

    The key points include six key points uniformly distributed on the boundary between the iris and the pupil, which are the first key point, the second key point, the third key point, the fourth key point, the fifth key point, and the sixth point. Key points; the key points also include four key points distributed on the boundary between the iris and the white of the eye, namely the 7th key point, the 8th key point, the 9th key point and the 10th key a point; wherein the first key point, the fourth key point, and the ninth key point are on a straight line, the second key point, the fifth key point, the seventh key point, and the tenth point The key points are on a straight line, and the third key point, the sixth key point, and the eighth key point are on a straight line.
18. The iris recognition device according to any one of claims 14 to 17, wherein

    The pre-processing module is specifically configured to perform polar coordinate transformation on the ROI to obtain a rectangular iris image; normalize the rectangular iris image to obtain the pre-processed iris image.
19. The iris recognition device according to claim 14, wherein

    The feature acquiring module is specifically configured to divide the preprocessed iris image into M image regions; the M is a positive integer greater than or equal to 2; and the 2D-Gabor filter is used in the M image regions. Each image area is convoluted to obtain a corresponding response amplitude; the response amplitude is encoded, and the code corresponding to the M image regions is combined to obtain the first set of feature data.
20. The iris recognition device according to claim 19, wherein

    The feature acquiring module is further specifically configured to perform convolution on each of the M image regions by using K frequency L direction 2D-Gabor filters to obtain K×L corresponding to each image region. Response amplitudes; binarizing the L response amplitudes at each frequency in each image region to obtain two codes corresponding to each frequency of each image region, combining each image region The K frequency obtains K×2 codes corresponding to each image region; and the code corresponding to the M image regions is combined to obtain the first group feature data, and the first group feature data includes M×K×2 codes.
21. The iris recognition device according to claim 20, wherein

    The feature acquisition module is further specifically configured to: when the nth response amplitude of the nth response amplitude The value is not greater than the n+1th response amplitude, the nth response amplitude is correspondingly binarized to 1, and when the nth response amplitude is greater than the n+1th response amplitude And the nth response amplitude is correspondingly binarized to 0, wherein the n is a positive integer greater than or equal to 1; combining the codes corresponding to the L response amplitudes to obtain L binarizations Encoding; obtaining two codes according to the L binarization codes, and combining K codes of each of the image regions to obtain K×2 codes.
22. The iris recognition device according to claim 14, wherein

    The feature acquiring module is further configured to divide the pre-processed iris image into N image regions, where N is a positive integer greater than or equal to 2; and acquiring each image region corresponding to the N image regions The LBP feature combines the LBP features corresponding to the N image regions to obtain the first set of feature vectors.
23. The iris recognition device according to claim 22, wherein

    The feature acquiring module is further specifically configured to perform binarization coding on each pixel in each image region to obtain a binarization code corresponding to each pixel; and according to the binarization code corresponding to all pixels in each image region. Obtaining a histogram corresponding to each image region; combining the histograms corresponding to the N image regions to obtain the first group of feature vectors.
24. The iris recognition device according to claim 23, wherein

    The feature acquiring module is further configured to: obtain a gray value of each pixel in each image region, and sequentially compare a gray value of each pixel with a gray value of eight neighbor pixels;

    When the gray value of the pixel point is greater than the gray value of the neighboring pixel point, the corresponding binarization code is 1; when the gray value of the pixel point is less than or equal to the gray value of the 8 neighborhood pixels, the corresponding two The valued code is 0, and the binarized code of the 8-bit byte corresponding to each pixel is obtained.
25. An iris recognition device according to claim 14 or 22 or 23 or 24, wherein

    The identification module is specifically configured to perform dimension reduction processing on the feature vector, and normalize the reduced dimension feature vector to obtain a normalized feature vector; calculate the normalized feature vector and the The vector distance of the pre-stored feature sequence.
26. The iris recognition device according to claim 14, wherein

    The feature acquisition module is further configured to divide the preprocessed iris image into H image regions, and process the H image regions by using 2D-Gabors of 1 frequency and J directions to obtain a second set of data features. ;

    The identification module is specifically configured to calculate a second Hamming distance between the second set of data features and the pre-stored feature data; calculate the first Hamming distance, the second Hamming distance, and the first A weighted value of a vector feature.
27. An iris recognition device, comprising:

    Processor and memory;

    The memory is used to store a program;

    The processor is configured to execute a program in the memory such that the iris recognition device performs the iris recognition method according to any one of claims 1 to 13.
28. A storage medium storing one or more programs, the instructions including instructions that, when executed by the iris recognition device including one or more processors, cause the iris recognition device to perform The iris recognition method according to any one of claims 1 to 13.

Images