CN114582008A - Living iris detection method based on two wave bands - Google Patents

Abstract

The invention discloses a dual-band-based living iris detection method belonging to the technical field of biological image processing. Including in vivo detection authority, perform in vivo iris detection, collect the user's eye region image I in the 530 nm band and the human eye region image II in the 700 nm band within the depth of field of the camera, and preprocess the human eye region image to obtain Grayscale iris image; obtain the tissue feature map I _f1 and blood vessel texture feature map I _f2 of the iris and sclera regions of interest; use the hash algorithm to calculate the difference between the tissue and blood vessel texture feature map of I and the tissue and blood vessel texture feature map of II According to the difference between the tissue feature maps I _f1 and II _f1 and the blood vessel texture feature maps I _f2 and II _f2 , the living body detection features are extracted, and the iris authenticity judgment result is output. The invention distinguishes the authenticity from the fake by the difference in absorption and reflectivity of the blood vessels and tissues in the living human eye and the biological texture simulated by the fake sample in different wavelength bands, and can well defend against the fake iris attack.

Description

A dual-band-based living iris detection method

technical field

The invention belongs to the technical field of biological image processing, and in particular relates to a dual-band-based living iris detection method.

Background technique

Iris recognition technology is similar in concept to fingerprint recognition technology, both of which are based on the unique physiological characteristics of the human body for identification. However, due to the acquisition equipment and methods of iris images, iris recognition requires a high degree of cooperation from the collector, and its friendliness is not as good as that of fingerprint recognition. Therefore, its commercialization degree has always been lower than that of fingerprint recognition and face recognition. In CN201910101762.3, in the patent for anti-counterfeiting method and device based on living iris, the user's first face image collected by a visible light camera is used to obtain the iris image of the user, and it is judged whether the iris image comes from the living iris, if the The iris image is derived from the living iris, and it is detected whether the iris image matches the iris image pre-stored by the user, and if it matches, the obtained face image is true; based on this, anti-counterfeiting judgment is performed. In today's digital information society, people will inevitably leave their physical features, fingerprint textures or images containing iris textures for various reasons. In this context, even if the iris recognition accuracy is high, if the attack of forged samples cannot be defended, once these physiological characteristics are obtained by criminals and used to steal property or information, it will bring about social disorder. To this end, the present invention proposes a dual-band-based living iris detection method to solve the problems mentioned in the above background art.

SUMMARY OF THE INVENTION

The object of the present invention is to provide a dual-band-based living iris detection method, characterized in that, two bands with the largest difference in the number of texture features of the living human eye and the fake sample are selected as the two control bands in the algorithm, The difference between the absorption and reflectivity in different wavelength bands of the intravascular tissue in the human eye and the biological texture simulated by the fake sample is used to distinguish the authenticity from the fake, including the following steps:

S1. The administrator enables the liveness detection permission in the background, and the user performs iris liveness detection on the client;

S2. The lighting module starts to work under the control of the central controller, and the user is within the working range of the lighting module and the depth of field of the camera;

S3. The central controller controls the light source to emit two sets of light with different wavelengths, namely light with wavelengths of 530 nm and 700 nm, so that the camera collects the user's eye region image I in the 530 nm band and human eye region image II in the 700 nm band respectively. , perform image preprocessing to reduce noise interference during positioning;

S4. Position the pupil, iris and sclera. First, use the adaptive threshold segmentation method to obtain a roughly segmented binary image of the pupil area according to the histogram of the iris image, and use the binary image with pupil position information to determine the center position of the pupil. The position of the center, set the iris area search range, and then use the Hough transform algorithm to accurately locate the iris area within the search range and segment the sclera area with less interference, and obtain the iris image I ₁ and the sclera image I ₂ ;

S5, operate on the iris image I ₁ and the sclera image I ₂ to obtain a tissue feature map I _f1 and a blood vessel texture feature map I _f2 of the region of interest;

S6, repeating steps S4-S5 for the preprocessed human eye region image II, to obtain a tissue feature map II _f1 and a blood vessel texture feature map II _f2 of the region of interest;

S7, extracting living body detection features based on the difference between the tissue feature maps I _f1 and II _f1 and the blood vessel texture feature maps I _f2 and II _f2 under the dual-band, inputting the living body features into the classifier for authenticity judgment, and outputting the iris authenticity judgment result.

In the step S3, the controller controls the light source of the camera to emit two sets of light with different wavelengths, namely light waves of 530 nanometers and 700 nanometers, so that the camera collects the user's human eye region image I in the 530 nanometer band and the human eye in the 700 nanometer band respectively. Regional image II, grayscale, Gaussian filtering and denoising are performed on the human eye region image, and nonlinear enhancement is performed on the filtered human eye region image to reduce noise interference such as eyelashes.

In the step S3, grayscale, Gaussian filtering and denoising are performed on the above-mentioned human eye region image, and the filtered human eye region image is subjected to a nonlinear enhancement operation to reduce eyelash interference, wherein I _g is to After filtering, the pixel in the 7*7 neighborhood of the filtered image takes the maximum value operation, and I _h is the minimum value operation for the pixel in the 7*7 neighborhood of I _g ;

In described step S4, use adaptive threshold segmentation method to carry out rough segmentation of pupil part, obtain binary image I _b (x, y) with pupil position information, determine pupil center in this binary image by formula (2) center of gravity method Position (C _x ,C _y ), where I _b (x,y) is the gray value at point (x,y), according to the position of the pupil center, set the search range, and use Hough transform within the search range The algorithm accurately locates the iris region and segmentes the sclera region with less interference, and obtains grayscale iris image I ₁ and grayscale scleral image I ₂ ;

In the step S5, the iris image and the sclera image obtained from the human eye region image I are operated by using the gradient enhancement-based blood vessel and tissue texture segmentation algorithm to obtain tissue and blood vessel texture feature maps.

In the step S6, the steps S4-S5 are repeated for the human eye region image II to obtain the tissue and blood vessel texture feature maps of the region of interest.

In the step S7, a hash algorithm is used to calculate the difference between the tissue and blood vessel texture feature map of I and the tissue and blood vessel texture feature map of II, and the difference is taken as the dual-band living iris detection feature of the present invention, and the feature is used to input classification The device performs authenticity judgment and outputs the iris authenticity judgment result.

In the step S5, the iris image I ₁ and the sclera image I ₂ are operated to obtain the tissue feature map I _f1 and the blood vessel texture feature map I _f2 of the region of interest. The general calculation method of the feature map I _fi (i=1,2) is as follows: 3) shown,

Where t(x, y) is the adaptive coefficient, k, b are empirical hyperparameters, here k=1, b=0.05, G _max (x, y) is I _i (i=1, 2) 5*5 The maximum gradient value in the window, G is the global maximum gradient value of the image, take G=255, m(x,y) is the average gray value of I _i (i=1,2)5*5 window, g(x,y ) is the local gradient at that point.

The beneficial effects of the present invention are as follows: the present invention takes the biological imaging characteristics of the iris and sclera of a living body under a dual-band light source as a starting point to conduct research and development of a living body detection algorithm, can resist a variety of forged sample attacks, has universality, and only needs to be The iris live detection can be completed by collecting two frames of images, which greatly shortens the detection time and is real-time. The feature of the invention is that the two bands with the largest difference in the texture characteristics of the living human eye and the fake sample are selected as the two control bands in the algorithm, and the absorption of the biological texture simulated by the intravascular tissue in the living human eye and the fake sample in different wavelength bands is used. The difference in reflectivity is used to distinguish the authenticity from the fake. This method can well defend against attacks such as printing remakes, colored contact lenses, and resin eyeball models.

Detailed ways

The invention provides a dual-band-based living iris detection method, which selects the two bands with the largest difference in the number of texture features of the living human eye and the fake sample as the two control bands in the algorithm. The difference between the absorption and reflectivity of the biological texture simulated by the fake sample in different wavelength bands is used to distinguish the authenticity from the fake. The present invention will be further described in detail below with reference to specific embodiments.

The dual-band-based living iris detection method comprises the following steps:

S1. The administrator enables the liveness detection permission in the background, and the user performs iris liveness detection on the client;

S2. The lighting module starts to work under the control of the central controller, and the user is within the working range of the lighting module and the depth of field of the camera;

S3. The central controller controls the light source of the camera to emit two sets of light with different wavelengths, namely 530 nm and 700 nm light, so that the camera collects the user's eye region image I in the 530 nm band and the human eye region image in the 700 nm band respectively. II. Perform grayscale, Gaussian filtering and denoising on the above-mentioned human eye region image, and perform nonlinear enhancement operation on the filtered human eye region image through formula (1) to reduce eyelash interference, where _Ig is the filtered image. The maximum value operation of the pixels in the 7*7 neighborhood, I _h is the minimum value operation for the pixels in the 7*7 neighborhood of I _g ;

S4, locate the pupil, iris and sclera, according to the distribution characteristics of the histogram of the iris image, use the adaptive threshold segmentation method to roughly segment the pupil part to obtain a binary image I _b (x, y) with pupil position information, through Formula (2) The center of gravity method determines the center position of the pupil (C _x ,C _y ) in the binary image, where I _b (x,y) is the gray value at the point (x,y), according to the position of the center of the pupil , set the search range, use the Hough transform algorithm to accurately locate the iris area and segment the sclera area with less interference within the search range, and obtain the grayscale iris image I ₁ and the grayscale sclera image I ₂ ;

S5. Operate the iris image I ₁ and the sclera image I ₂ to obtain the tissue feature map I _f1 and the blood vessel texture feature map I _f2 of the region of interest. The general calculation method of the feature map I _fi (i=1, 2) is as shown in formula (3) shown, where t(x, y) is the adaptive coefficient, k, b are empirical hyperparameters, here k=1, b=0.05, G _max (x, y) is I _i (i=1, 2) The maximum gradient value in the 5*5 window, G is the global maximum gradient value of the image, take G=255, m(x,y) is the average gray value of I _i (i=1,2) in the 5*5 window, g( x, y) is the local gradient at the point;

S6, repeating steps S4-S5 for the human eye region image II, to obtain the tissue feature map II _f1 of the region of interest and the blood vessel texture feature map II _f2 ;

S7. Calculate the difference between the tissue feature map I _f1 and II _f1 , and the difference between the blood vessel texture feature map I _f2 and II _f2 by using a hash algorithm, use the difference as a dual-band living iris detection feature, and use the feature to input the classifier to carry out Authenticity judgment, output the iris authenticity judgment result, taking the difference between the feature map I _f1 and II _f1 as an example, the pseudo code of the hash algorithm is shown in Table 1.

Table 1 Pseudo code of hash algorithm

Specifically, the specific steps of collecting the human eye region image I and the human eye region image II in the step S3 include: the controller controls the light source of the camera to emit two sets of light with different wavelengths, that is, light waves of 530 nanometers and 700 nanometers, so that The camera collects the user's eye region image I in the 530 nm band and the human eye region image II in the 700 nm band respectively, performs grayscale, Gaussian filtering and denoising on the human eye region image, and performs a non-decoding process on the filtered human eye region image. Linear enhancement operation to reduce noise interference such as eyelashes.

Specifically, the specific steps of obtaining the tissue and blood vessel texture feature map in the step S5 include: using a gradient enhancement-based blood vessel and tissue texture segmentation algorithm to operate the iris image and sclera image obtained from the human eye region image I to obtain the tissue and blood vessels. Vessel texture feature map.

Specifically, in the step S6, steps S4-S5 are repeated for the human eye region image II to obtain the tissue and blood vessel texture feature maps of the region of interest.

Specifically, the specific steps of extracting and judging the features of the living body detection in the step S7 include: using a hash algorithm to calculate the degree of difference between the tissue and blood vessel texture feature map of I and the tissue and blood vessel texture feature map of II, and this degree of difference is taken as this paper. The dual-band living iris detection feature is used to input the classifier for authenticity judgment, and the iris authenticity judgment result is output.

To sum up, the present invention provides a dual-band-based living iris detection method. Compared with other methods, the present invention takes the biological imaging characteristics of the living iris and sclera under the dual-band light source as a starting point to conduct research on the living detection algorithm. And developed, can resist a variety of forged sample attacks, has universality, and only needs to collect two frames of images to complete the living iris detection, greatly shortening the detection time, and has real-time performance.

The present invention selects two wavebands with the largest difference in the number of texture features of the living human eye and the fake sample as the two control wavebands in the algorithm. Differences to distinguish between true and false. This method can well defend against fake iris attacks.

Claims (8)

1. a living body iris detection method based on dual wavebands, is characterized in that, choose living body human eye and two wavebands that forged sample texture feature quantity variation difference is the largest as two contrast wavebands in the algorithm, pass through the blood vessel in living body human eye. The difference between the absorption and reflectivity of the biological texture simulated by the tissue and the fake sample in different wavelength bands is used to distinguish the authenticity from the fake, including the following steps: S1. The administrator enables the liveness detection permission in the background, and the user performs iris liveness detection on the client; S2. The lighting module starts to work under the control of the central controller, and the user is within the working range of the lighting module and the depth of field of the camera; S3. The central controller controls the light source to emit two sets of light with different wavelengths, namely light with wavelengths of 530 nm and 700 nm, so that the camera collects the user's eye region image I in the 530 nm band and human eye region image II in the 700 nm band respectively. , perform image preprocessing to reduce noise interference during positioning; S4. Position the pupil, iris and sclera. First, use the adaptive threshold segmentation method to obtain a roughly segmented binary image of the pupil area according to the histogram of the iris image, and use the binary image with pupil position information to determine the center position of the pupil. The position of the center, set the iris area search range, and then use the Hough transform algorithm to accurately locate the iris area within the search range and segment the sclera area with less interference, and obtain the iris image I ₁ and the sclera image I ₂ ; S5, operate on the iris image I ₁ and the sclera image I ₂ to obtain a tissue feature map I _f1 and a blood vessel texture feature map I _f2 of the region of interest; S6, repeating steps S4-S5 for the preprocessed human eye region image II, to obtain a tissue feature map II _f1 and a blood vessel texture feature map II _f2 of the region of interest; S7, extracting living body detection features based on the difference between the tissue feature maps I _f1 and II _f1 and the blood vessel texture feature maps I _f2 and II _f2 under the dual-band, inputting the living body features into the classifier for authenticity judgment, and outputting the iris authenticity judgment result. 2. the living iris detection method based on dual wavebands according to claim 1, is characterized in that, in described step S3, the light source of controller control camera emits the light of two groups of different wavelengths respectively, namely the light waves of 530 nanometers and 700 nanometers, Make the camera collect the user's eye region image I in the 530 nm band and the human eye region image II in the 700 nm band respectively, perform grayscale, Gaussian filtering and denoising on the human eye region image, and perform non-decoding on the filtered human eye region image. Linear enhancement operation to reduce noise interference such as eyelashes. 3. The method for detecting a living iris based on dual wavebands according to claim 2, wherein in the step S3, grayscale, Gaussian filtering and denoising are performed on the above-mentioned human eye region image, and the filtered images are processed by formula (1). The non-linear enhancement operation is performed on the image of the human eye region to reduce eyelash interference, where I _g is the maximum value operation for the pixels in the 7*7 neighborhood of the filtered image, and I _h is the minimum value for the pixels in the 7*7 neighborhood of the _Ig operate;

4. the living iris detection method based on dual wavebands according to claim 1, is characterized in that, in described step S4, uses adaptive threshold segmentation method to carry out rough segmentation to pupil part, obtains the binary image I _b with pupil position information (x, y), the center position of the pupil (C _x ,C _y ) in the binary image is determined by the barycentric method of formula (2), where I _b (x, y) is the gray value at the point (x, y) , according to the position of the pupil center, set the search range, within the search range, use the Hough transform algorithm to accurately locate the iris area and segment the sclera area with less interference, and obtain the grayscale iris image I ₁ and the grayscale sclera Image I ₂ ;

5. the living iris detection method based on dual wavebands according to claim 1, is characterized in that, utilizes the blood vessel and tissue texture segmentation algorithm based on gradient enhancement in described step S5, to the iris image and the iris image obtained from human eye region image 1. Scleral image manipulation to obtain tissue and vessel texture feature maps. 6 . The dual-band-based living iris detection method according to claim 1 , wherein steps S4 - S5 are repeated for the human eye region image II in the step S6 to obtain the tissue and blood vessel texture feature maps of the region of interest. 7 . 7. The dual-band-based living iris detection method according to claim 1, wherein in the step S7, a hash algorithm is used to calculate the difference between the tissue and blood vessel texture feature map of I and the tissue and blood vessel texture feature map of II The degree of difference is taken as the dual-band living iris detection feature of the present invention, and the feature is used to input the classifier for authenticity judgment, and the iris authenticity judgment result is output. 8. The dual-band-based living iris detection method according to claim 1, wherein in the step S5, the iris image I ₁ and the sclera image I ₂ are operated to obtain the tissue feature map I _f1 of the region of interest and the texture features of blood vessels Figure I _f2 , the general calculation method of the feature map I _fi (i=1,2) is shown in formula (3),

Where t(x, y) is the adaptive coefficient, k, b are empirical hyperparameters, here k=1, b=0.05, G _max (x, y) is I _i (i=1, 2) 5*5 The maximum gradient value in the window, G is the global maximum gradient value of the image, take G=255, m(x,y) is the average gray value of I _i (i=1,2)5*5 window, g(x,y ) is the local gradient at that point.