Biometric-based mobile phones

Abstract

The present application relates to a mobile phone based on biometrics, which includes a mobile phone body, a camera device and a processor. The camera device is installed on the mobile phone body and is used to obtain external image information. The processor can perform the following steps: obtain image information, and pre-process the image information to obtain an iris feature image; standardize and correct the iris feature image, and obtain the iris center; input the iris feature image and the iris center into a convolutional neural network recognition model to obtain radial pattern information and radial pattern intersection position information of the iris feature image, wherein the radial pattern information includes position information and angle information, and the convolutional neural network can use the iris center as the convolution starting point and perform convolution along the radial lines from the iris center to the outside; match the radial pattern information and the intersection position with the pre-stored information in the database, so as to obtain the corresponding identity. The present application has the effect of realizing fast and accurate iris recognition.

Description

Technical Field

[0001] The present application relates to the field of biometrics, and in particular to a mobile phone based on biometrics. Background Art

[0002] With the rapid development of mobile communication technology, mobile phones have become an indispensable part of people's daily lives. Smartphones not only provide communication functions, but also become important tools for people's work, entertainment, social interaction, etc. The ubiquity of mobile phones means their potential in many emerging application areas, including security verification and identity recognition.

[0003] As people's awareness of privacy protection gradually increases, traditional authentication methods such as passwords and fingerprints have gradually exposed some security loopholes. Therefore, more advanced and secure identity authentication methods have received widespread attention. As a biometric technology, iris recognition has become a highly potential means of identity authentication due to its uniqueness and stability.

[0004] Currently, many smartphones are equipped with advanced front cameras that can easily obtain users' iris images, providing a hardware foundation for iris recognition. However, the challenge that comes with it is that users have very high requirements for the immediacy of identity authentication processes such as unlocking their phones, which poses a strict challenge to the processing speed of iris images.

[0005] However, the computing power of mobile phones is relatively limited, especially compared with data centers and dedicated servers, where computing resources are more scarce. Iris image processing usually uses advanced algorithms such as deep neural networks and convolutional neural networks. The complex nonlinear activation functions and a large number of multi-term calculations in these algorithms have extremely high requirements for computing power. If these algorithms are simplified to adapt to the computing power and response time of mobile phones, the recognition effect may be significantly reduced.

[0006] Therefore, the current challenge is to find a fast and accurate iris recognition technology while ensuring the accuracy and security of iris recognition and taking into account the computing power limitations of mobile phones and the high response time requirements of users. This is not only a technical issue, but also involves a wide range of user experience and commercial applications. Summary of the invention

[0007] In order to achieve fast and accurate iris recognition, the present application provides a mobile phone based on biometrics.

[0008] In the first aspect, the present application provides a mobile phone based on biometrics, which adopts the following technical solution:

[0009] A mobile phone based on biometrics, comprising a mobile phone body, a camera device and a processor, wherein the camera device is installed on the mobile phone body and used to obtain image information of the outside world, and the processor can perform the following steps:

[0010] S1. Obtain image information and preprocess the image information to obtain an iris feature image;

[0011] S2. performing standardization correction on the iris feature image and obtaining the iris center;

[0012] S3. Inputting the iris feature image and the iris center into the convolutional neural network recognition model to obtain the radial pattern information and the radial pattern intersection position information of the iris feature image, wherein the radial pattern information includes the position information and the angle information, and the convolutional neural network can take the iris center as the convolution starting point and perform convolution along the radial lines outward from the iris center;

[0013] S4. Match the radial pattern information and the intersection position with the pre-stored information in the database to obtain the corresponding identity.

[0014] By adopting the above technical solution, iris image information is obtained through the front camera and / or iris camera of the mobile phone, and processed into a standard style to adapt to the convolutional neural network recognition model, and then the radial pattern information and the position information of the intersection of the radial pattern of the iris feature image are obtained by processing through convolution and other methods. Since the convolutional neural network takes the center of the iris as the starting point of the convolution and performs convolution along the radial lines from the center of the iris to the outside, it is adapted to the feature that the iris of some people has a radial texture. Therefore, this type of convolutional neural network can directly use a linear activation function, and at the same time, the generalization requirements of the model are low, so the requirements for the computing power of the mobile phone are low, and it can respond quickly.

[0015] Optionally, S1 includes the following sub-steps:

[0016] S11. Obtain an image to be tested, and input the image to be tested into a target detection model, determine a target bounding box based on target box detection, and determine a target individual set as a contour area based on semantic segmentation; wherein the target is an iris, the target bounding box is used to reflect the imaging position of the iris in the image to be tested, the target individual set is used to reflect the pixel points located in the iris imaging area, and the target detection model is a pre-trained instance segmentation model;

[0017] S12. Segmenting the image to be tested into images each containing an iris based on the target bounding box;

[0018] S13. Based on the pixel information of adjacent pixels of the image to be tested, pixel interpolation is performed between adjacent pixels to enlarge the sample image;

[0019] S14. Combine the target external frame and the target individual set to output an iris feature image; wherein the iris feature image is pixel information of pixel points.

[0020] By adopting the above technical solution, through the target detection and image segmentation of S11 and S12, the iris in the image to be tested is accurately located and segmented, ensuring the accurate acquisition of the iris. The pre-trained instance segmentation model can efficiently identify the position of the iris in the image, thereby increasing the recognition accuracy. The pixel interpolation step of S13 can enlarge the sample image, increase the resolution of the image, and facilitate further iris feature extraction. This step can effectively improve the accuracy and stability of subsequent iris feature analysis. S14 combines the target external frame and the target individual set, and outputs the iris feature image as pixel information of pixel points. This step converts the characteristics of the iris into a data form that can be further analyzed, providing an effective basis for subsequent iris recognition and standardization.

[0021] Optionally, S2 includes the following sub-steps:

[0022] S21. Perform instance segmentation on the iris feature image to obtain a preliminary screening instance of the iris;

[0023] S22. Extracting iris edge information from the iris feature image, locating the center of the inner and outer iris circles, and calculating geometric parameters; wherein the geometric parameters are the radius, or the major axis and the minor axis;

[0024] S23. Calculate the center coordinate deviation of the inner circle and the outer circle, and then determine the physiological deviation or sampling deviation based on the deviation;

[0025] S24. Standardizing the iris feature image based on deviation type and geometric shape classification;

[0026] S25. Calculate the ratio of the inner and outer iris radii as a measure of the degree of pupil scaling;

[0027] S26. Using a correction function, taking the radius ratio of the inner and outer edges of the iris as a parameter, radial nonlinear sampling is performed to correct the uneven scaling of the pupil.

[0028] By adopting the above technical solution, the geometric characteristics of the iris can be accurately described through instance segmentation and extraction of iris edge information. In particular, by locating the center of the inner and outer circles of the iris and calculating geometric parameters (such as radius, major axis and minor axis), the structural characteristics of the iris can be described in detail. By calculating the center coordinate deviation of the inner and outer circles, it is further determined whether it is a physiological deviation or a sampling deviation. This intelligent analysis can effectively distinguish between normal changes in iris images and potential sampling errors, thereby providing an accurate basis for subsequent iris standardization. By calculating the radius ratio of the inner and outer circles of the iris, this step can accurately measure the degree of zoom of the pupil of the eye. This measurement not only helps to understand the physiological state of the iris, but also provides important reference information for further analysis of iris features.

[0029] The above steps provide a flexible and universal way to process iris images. Whether it is the shape, size or position of the iris, these steps can be used to accurately process and adjust it, which enhances the versatility and adaptability of the system.

[0030] Optionally, the S23 includes the following sub-steps:

[0031] S231. Calculate the center coordinate deviation of the inner circle and the outer circle;

[0032] S232. Determine the type of deviation;

[0033] S233. Analyze and classify the causes of deviations;

[0034] S234. Adjust iris positioning according to the deviation;

[0035] S235. Store the adjusted iris information.

[0036] By adopting the above technical solution, the coordinate deviation of the center of the inner and outer circles of the iris is calculated, and the type of deviation is further determined. This meticulous calculation and judgment process ensures the initial accuracy of iris recognition and provides an accurate basis for subsequent analysis and adjustment. By analyzing the causes of the deviation and classifying them, this step can accurately identify abnormalities and changes in iris images, such as physiological deviations or sampling deviations.

[0037] Optionally, the S24 includes the following sub-steps:

[0038] S241. If the iris feature image is an ellipse, obtain the geometric parameters of the ellipse iris and calculate the transformation ratio from the ellipse to the circle;

[0039] S242. Mapping the ellipse to the circle based on the affine transformation, and interpolating the transformed iris feature image based on the transformation ratio;

[0040] S243. Calibrate the transformed iris feature image to ensure accuracy.

[0041] By adopting the above technical solution, this step converts the iris' elliptical shape into a standard circle, thereby simplifying the subsequent processing process. At the same time, by calibrating the transformed iris feature image, the accuracy and consistency of the image are ensured. This precise calibration process not only improves the accuracy of iris recognition, but also enhances the system's adaptability to various iris shapes and conditions.

[0042] Optionally, S3 includes the following sub-steps:

[0043] S31. Obtaining the iris center position based on the corrected iris feature image, and establishing a polar coordinate system based on the iris center position;

[0044] S32. Based on the iris center position, a plurality of radial sampling paths with equal angular spacing are sequentially set in a circular direction around the iris center, and a circular sampling path is set outward along the radial direction of the radial sampling path based on the iris center position;

[0045] S33. Sampling the iris feature image based on the intersection of each annular sampling path and the radial sampling path;

[0046] S34. Rearrange the obtained sampling point information into a new feature image, wherein each row of pixel points of the new feature image corresponds to each sampling point on the same radial sampling path on the original iris feature image and has the same order, and each column of pixel points corresponds to each sampling point on the same circumferential sampling path on the original iris feature image and has the same order;

[0047] S35. Input the new feature image into the convolutional neural network and perform a convolution operation for filtering;

[0048] S36. Set the activation function and perform the maximum pooling operation;

[0049] S37. Select a convolution kernel for detecting horizontal lines and a convolution kernel for detecting intersections and perform convolution operations on the pooled images to obtain spacing features and intersection features;

[0050] S38. Perform an inverse transformation from the rectangular coordinate system to the polar coordinate system based on the spacing features and the intersection features to obtain the radial pattern information and the radial pattern intersection position information in the iris feature image.

[0051] By adopting the above technical solution, since the eyes of some people have texture features that diverge from the pupil to the outside, and the partial bending deviation of the texture features produces intersections, for this type of iris, such features can be used as a priori conditions for independent judgment. Since there may be multiple irises in the image, it is necessary to perform instance segmentation and select the target iris. Based on the target iris image after instance segmentation, the center position is selected, and then a polar coordinate system is established based on the center position. Based on the center position of the iris, several radial sampling paths with equal angular spacing are sequentially set in a circular direction around the center of the iris, and a circular sampling path is set outward along the radial direction of the radial sampling path based on the center position of the iris. In this way, the radial sampling path is easy to correspond to the texture of the iris, so that it can be processed continuously during convolution, improving the efficiency of feature recognition. Then mapping is performed so that the pixel points of each row of the new image correspond to the pixel points on the same radial line on the old image. In this way, since the convolution kernel of the convolutional neural network usually moves row by row along the convolved matrix, the stripe features on the new image can be better recognized.

[0052] Optionally, S4 includes the following steps:

[0053] S41. Loading pre-stored information from a database, wherein the pre-stored information database includes a series of iris radial pattern information and intersection locations of known identities;

[0054] S42. Compare the newly acquired iris feature information with the iris features in the database and calculate the similarity or distance;

[0055] S43. Based on the comparison result, the iris feature in the database that is most similar to or closest to the new iris feature information is matched, and the corresponding identity tag is retrieved;

[0056] S44. combining the retrieved identity tag with other identity verification information to improve the security of identification;

[0057] S45. After confirming the identity, perform corresponding operations according to the application scenario.

[0058] By adopting the above technical solution, by loading the pre-stored information database including the iris radial pattern information and intersection position of the known identity, this step ensures the rapid start and high efficiency of the iris recognition process, and also provides an accurate reference basis for the subsequent feature comparison. This step compares the newly acquired iris feature information with the iris features in the database and calculates the similarity or distance. Based on the comparison result, the system can match the database iris feature that is most similar or closest to the new iris feature information and retrieve its corresponding identity tag. This process enhances the accuracy and reliability of iris recognition and ensures the security of the system. By combining the retrieved identity tag with other authentication information, the security of identification is enhanced. This multi-factor authentication mechanism provides an additional layer of protection, further improving the security level of the entire system. After confirming the identity, the system can perform corresponding operations according to the application scenario. This flexible automatic execution function not only increases the convenience of users, but also allows the system to adapt to a variety of different application scenarios and needs.

[0059] In a second aspect, the present application provides a computer device, which adopts the following technical solution:

[0060] A computer device comprising:

[0061] one or more processors;

[0062] Memory;

[0063] One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to execute the steps performed by the processor of the mobile phone.

[0064] In a third aspect, the present application provides a computer-readable storage medium, which adopts the following technical solution:

[0065] A computer-readable storage medium stores a computer program that can be loaded by a processor and execute the above method.

[0066] The storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set are loaded and executed by the processor to implement the steps performed by the processor of the mobile phone as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 A flowchart of steps executed by a processor in a mobile phone according to an embodiment of the present invention is shown.

[0068] Figure 2 A flowchart of sub-step S1 in one embodiment of the present invention is shown.

[0069] Figure 3 A flowchart of sub-step S2 in one embodiment of the present invention is shown.

[0070] Figure 4 A flowchart of sub-step S23 in one embodiment of the present invention is shown.

[0071] Figure 5 A flowchart of sub-step S24 in one embodiment of the present invention is shown.

[0072] Figure 6 A flowchart of sub-step S3 in one embodiment of the present invention is shown.

[0073] Figure 7 A flowchart of sub-step S4 in one embodiment of the present invention is shown. DETAILED DESCRIPTION

[0074] The present application is further described in detail below in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0075] In the following description, for the purpose of explanation, many specific details are set forth in order to provide a thorough understanding of the inventive concept. As part of this specification, some of the drawings in the drawings of the present disclosure represent structures and devices in the form of block diagrams to avoid making the disclosed principles complicated and difficult to understand. For the sake of clarity, not all features of the actual implementation are necessary to be described. In addition, the language used in this disclosure has been selected primarily for readability and instructional purposes, and may not be selected to delineate or limit the subject matter of the present invention, thereby resorting to the necessary claims to determine such inventive subject matter. References to "a specific implementation" or "specific implementations" in this disclosure mean that the specific features, structures or characteristics described in conjunction with the specific implementation are included in at least one specific implementation, and multiple references to "a specific implementation" or "specific implementations" should not be understood as necessarily referring to the same specific implementation.

[0076] Unless expressly limited, the terms "a", "an", and "the" are not intended to refer to a singular entity, but rather to include general categories of which specific examples may be used for illustration. Thus, the use of the terms "a" or "an" may mean any number of at least one, including "one", "one or more", "at least one", and "one or more than one". The term "or" means any of the alternatives and any combination of the alternatives, including all alternatives, unless the alternatives are expressly indicated to be mutually exclusive. The phrase "at least one of" when combined with a list of items refers to a single item in the list or any combination of the items in the list. The phrase does not require all of the listed items unless expressly limited to that.

[0077] With the rapid advancement of technology, mobile phone performance has been improved unprecedentedly. The integration of advanced processors, large storage space, high-resolution display screens and other technologies has made modern smartphones powerful portable computing devices. At the same time, people's awareness of privacy protection is also increasing, and the security requirements for personal information are also getting higher and higher.

[0078] In this context, the front camera and iris camera in mobile phones have begun to play an important role in iris recognition technology. Using the unique texture of the iris for identity authentication not only provides a high degree of security, but also brings a convenient user experience. Iris recognition technology is also increasingly closely associated with deep learning networks and convolutional neural networks. Through complex neural network algorithms, iris features can be analyzed and identified more accurately, thereby further improving the accuracy and reliability of recognition.

[0079] However, the response speed of the iris recognition system may be affected by many factors, such as the quality of iris image acquisition, the complexity of the processing algorithm, the limitation of hardware performance, etc. Response speed is one of the key factors in the practicality of the iris recognition system and has a direct impact on the user experience. Therefore, how to optimize the response speed while ensuring recognition accuracy so that iris recognition can better meet the needs of real-time and efficient applications has become an important topic in the current development of iris recognition technology.

[0080] The development of modern smartphones has promoted the advancement and popularization of iris recognition technology, and the application of deep learning and convolutional neural networks has brought new breakthroughs to iris recognition. Deep learning is an algorithm based on artificial neural networks that allows computers to recognize complex patterns and features by learning from a large number of samples. In iris recognition, deep learning can be used to automatically identify and analyze the unique texture and structural features of the iris. Convolutional neural networks (CNNs) are a special form of deep learning that are particularly suitable for processing image data. In iris recognition, convolutional neural networks can be used to automatically extract key features from iris images. Specifically, CNNs can gradually extract low-level to high-level features of an image through multiple convolutional layers, activation layers, and pooling layers.

[0081] For example, through the multi-layer structure of CNN, complex features such as edges, textures, and shapes can be extracted from the original iris image, which are critical for iris recognition. Using supervised training of deep learning, the model can learn to recognize the iris features of different individuals. The trained model can automatically classify unknown iris images, thereby achieving accurate identity verification. Deep learning can increase the diversity of training samples through data augmentation techniques, such as rotating, scaling, and shearing iris images, thereby enhancing the generalization ability of the model.

[0082] However, the computing power of mobile phones is relatively limited, especially compared with data centers and dedicated servers, where computing resources are more scarce. Since user iris image processing usually uses advanced algorithms such as deep neural networks and convolutional neural networks, the complex nonlinear activation functions and a large number of multiple terms in these algorithms have extremely high requirements for computing power. However, in order to improve the user experience, the current unlocking of mobile phone screens has reached the millisecond level. If these algorithms are simplified to adapt to the computing power and response time of mobile phones, the recognition effect may be significantly reduced.

[0083] Therefore, the present application proposes a mobile phone based on biometrics, which includes a mobile phone body, a camera device and a processor. The camera device is installed on the mobile phone body and is used to obtain image information from the outside world. The mobile phone body is used to cooperate with the processor to realize the display, calculation, networking and other functions of the mobile phone. In other words, the mobile phone body and the processor can cooperate to realize most functions except the camera function.

[0084] The camera device includes a front camera and an iris camera. The front camera is located on the front of the phone and is usually used for selfies, video calls, facial recognition and other functions. The resolution of the front camera is generally low, but with the advancement of technology, the front cameras of many modern smartphones are already able to provide HD or even UHD image quality. The main function of the iris camera is to capture iris images. The iris is the colored part of the eye, and the iris texture of each person is unique. The iris camera can accurately capture these textures for identity verification. Many iris cameras use infrared light sources to illuminate the iris. Infrared light can better highlight the texture of the iris and provide stable images under different lighting conditions.

[0085] Iris cameras capture images of the iris through a specialized lens system. These lenses are specially designed to capture the minute details of the iris. An infrared light source illuminates the iris, highlighting its texture and reducing interference from ambient light. Iris cameras may include built-in image processing capabilities to pre-process the iris image. Pre-processing may include noise reduction, sharpening, and contrast adjustment for further analysis.

[0086] Reference Figure 1 , based on the camera device, the processor can perform the following steps:

[0087] S1. Acquire image information, and preprocess the image information to obtain an iris feature image.

[0088] Specifically, refer to Figure 2 In one embodiment, S1 includes the following sub-steps:

[0089] S11. Obtain an image to be tested, and input the image to be tested into a target detection model, determine a target bounding box based on target box detection, and determine a target individual set as a contour area based on semantic segmentation; wherein the target is an iris, the target bounding box is used to reflect the imaging position of the iris in the image to be tested, the target individual set is used to reflect the pixel points located in the iris imaging area, and the target detection model is a pre-trained instance segmentation model.

[0090] S12. Segment the image to be measured into images each containing an iris based on the target bounding box.

[0091] For example, an iris camera captures a high-resolution image of the user's eye, which includes various eye structures such as the iris, pupil, sclera, etc. For example, the image may have a resolution of 1280 x 720 and accurately capture the color and texture of the iris.

[0092] The captured image is then fed into a pre-trained instance segmentation model. This model can be deep learning based such as Mask R-CNN or YOLO etc. This model has been trained to recognize and localize irises in images.

[0093] The model analyzes the image, determines the approximate location of the iris, and then draws a bounding box around the outline of the iris. The bounding box reflects the approximate location and size of the iris in the image. For example, the bounding box may include the coordinates (x, y) of the center of the iris, as well as the width and height of the iris.

[0094] In addition to the object bounding box, the system further determines the precise outline of the iris through semantic segmentation. Semantic segmentation involves classifying each pixel, so it can identify each pixel within the iris imaging area in the image. This helps distinguish the iris from other eye structures, such as the pupil and sclera.

[0095] Through semantic segmentation, we obtain a target individual set, which includes all the pixels in the iris area. This set reflects the precise shape of the iris contour and can be used for further analysis and processing.

[0096] S13. Based on the pixel information of adjacent pixels of the image to be tested, pixel interpolation is performed between adjacent pixels to enlarge the sample image.

[0097] When enlarging an image, you need to choose a pixel interpolation method. Interpolation methods can be nearest neighbor, bilinear, bicubic, etc. Each method has its own characteristics and applicable scenarios. Based on the selected interpolation method, new pixel values ​​between adjacent pixels can be calculated. For example, in bilinear interpolation, the value of the new pixel is a weighted average of the values ​​of the four surrounding pixels. Using the values ​​of the interpolated pixels, a new, larger iris image needs to be created. This new image includes the original pixels and the interpolated pixels to form a higher resolution image.

[0098] For example, for the nearest neighbor interpolation, the nearest neighbor interpolation is the simplest interpolation method, which only needs to take the gray value of the known pixel closest to the target point as the gray value of the target point. Its advantages are simple calculation and fast speed. The disadvantage is that it may cause image jagged phenomenon and poor image quality.

[0099] Bilinear interpolation is a linear interpolation in two directions. The gray value of the target point is calculated by the gray values ​​of the four surrounding known points and their distances from the target point. Its advantages are that it is relatively simple and the image quality is better than the nearest neighbor interpolation. Its disadvantage is that it may introduce some blur.

[0100] Bicubic interpolation is a cubic polynomial interpolation in two directions, usually using 16 known points around for calculation. Its advantages are good image quality, smooth edges, and good detail preservation. Its disadvantage is that the calculation complexity is higher than the previous two.

[0101] S14. Combine the target external frame and the target individual set to output an iris feature image; wherein the iris feature image is pixel information of pixel points.

[0102] An appropriate feature extraction method is selected to capture the unique features of the iris. One or more feature extraction algorithms such as Gabor filter, Laplacian filter, etc. can be used. Based on the target bounding box and the set of target individuals, the precise location and boundaries of the iris are determined. This allows accurate focus on the iris and ignores other irrelevant image areas. Using the selected feature extraction method, specific calculations and operations are applied to the focused iris area to form an iris feature image. This feature image captures the unique pattern and structure of the iris. The feature image can be saved as a file for future comparison or analysis, or directly passed to the next step for real-time processing.

[0103] S2. Standardize and correct the iris feature image and obtain the iris center.

[0104] Specifically, refer to Figure 3 In one embodiment, S2 includes the following sub-steps:

[0105] S21. Perform instance segmentation on the iris feature image to obtain preliminary screening instances of the iris.

[0106] S22. Extract iris edge information from the iris feature image, locate the center of the inner circle and the outer circle of the iris, and calculate geometric parameters; wherein the geometric parameters are the radius, or the major axis and the minor axis.

[0107] S23. Calculate the center coordinate deviation between the inner circle and the outer circle, and then determine the physiological deviation or sampling deviation based on the deviation.

[0108] Specifically, refer to Figure 4 In one embodiment, S23 includes the following sub-steps:

[0109] S231. Calculate the center coordinate deviation of the inner circle and the outer circle.

[0110] First, obtain the center coordinates of the inner and outer circles of the iris, such as the inner circle center C i =(x i ,y i ), outer ring center C o =x o ,y o ). Calculate the coordinate deviation as Δx = x i -x o ,Δy=y i -y o For the iris of the human eye, this deviation should reflect the normal physiological state or the distortion during the shooting process.

[0111] S232. Determine the type of deviation.

[0112] A threshold value, such as 3-5 pixels, is set based on the resolution of the iris feature image and prior knowledge. This threshold value is used to determine the source of the deviation between the inner and outer centers of the iris, that is, to determine whether it is a physiological deviation or a deviation caused by eye movement during acquisition. The specific algorithm can be used threshold to determine.

[0113] S233. Analyze and classify the causes of deviations.

[0114] For physiological deviation, it can be further analyzed that it is related to the special physiological structure of the iris or certain ophthalmic diseases. For sampling deviation, it may be related to factors such as lighting, angle, and reflection during the sampling process.

[0115] S234. Adjust iris positioning according to the deviation.

[0116] Depending on the type and size of the deviation identified, different calibration algorithms can be used to adjust the iris positioning. For example, if it is identified as a physiological deviation, it can be corrected based on the general elliptical nature of the iris shape. If it is a deviation during acquisition, it may be necessary to adjust the parameters in the iris positioning algorithm or consider specific factors in the acquisition process, such as light source, angle, etc.

[0117] S235. Store the adjusted iris information.

[0118] The center coordinates, radius and other information of the corrected iris inner and outer circles are stored for use in subsequent steps. This information can be saved in a specific data structure for easy retrieval and use in subsequent steps.

[0119] S24. Normalize the iris feature image based on deviation type and geometric shape classification.

[0120] Specifically, refer to Figure 5 In one embodiment, S24 includes the following sub-steps:

[0121] S241. If the iris feature image is an ellipse, obtain geometric parameters of the ellipse iris and calculate the transformation ratio from the ellipse to the circle.

[0122] An edge detection algorithm, such as the Canny algorithm, is used to determine the outer boundary of the iris, and the parameters of the ellipse are further fitted using the least squares method to obtain the major and minor axis lengths of the iris.

[0123] S242. Map the ellipse to the circle based on the affine transformation, and interpolate the transformed iris feature image based on the transformation ratio.

[0124] Determine the center coordinates (x0, y0), major axis length a, and minor axis length b of the ellipse. These parameters can be obtained from the original image through edge detection and ellipse fitting algorithms. Since the elliptical iris is to be converted into a standard circle, it is necessary to calculate the transformation ratio, which is equal to the ratio between the major axis length a and the minor axis length b. For example, if a = 60 pixels and b = 40 pixels, the transformation ratio is 60 / 40 = 1.5. The ellipse can be converted into a circle by applying a linear or nonlinear transformation, such as an affine transformation or a polar coordinate transformation. The specific algorithm may involve steps such as coordinate transformation and interpolation. If the ratio of the major axis to the minor axis is 1.5, the elliptical iris can be converted into a circle by stretching it 1.5 times in the direction of the minor axis of the ellipse. This can ensure the quality of the converted image through existing interpolation techniques.

[0125] S243. Calibrate the transformed iris feature image to ensure accuracy.

[0126] Since the transformation may introduce some distortion or distortion, further adjustments and calibration may be required, such as realigning the image center and ensuring that the transformed circular iris is perfectly aligned with the standard size.

[0127] S25. Calculate the ratio of the inner and outer iris radii as a measure of the degree of pupil scaling.

[0128] Since the resolution of the image obtained during each iris feature image acquisition process may be different, it may not be possible to directly compare the iris feature images. In particular, the resolution difference of iris feature images obtained by different devices, different environmental conditions and different acquisition times may be further increased. Therefore, it is inappropriate to use an absolute scale such as pupil diameter to measure the size of the pupil in an image because it is not universally applicable and comparable.

[0129] Currently, measurement and statistical research in ophthalmology provides a solution. A large amount of statistical data and experimental analysis show that the iris diameter of adult eyes varies little between individuals and can be considered almost consistent. This discovery provides a key reference standard for iris recognition, that is, although the resolution of iris feature images may vary, the iris diameter can serve as a stable benchmark.

[0130] Therefore, the ratio of pupil diameter to iris diameter can be used as a measure of pupil scaling. This ratio not only eliminates the impact of resolution, but also provides a relatively stable and accurate measurement method. Compared with absolute pupil size, this relative size measurement is more robust and can provide a consistent measurement standard in different scenes and conditions.

[0131] Hough transform can detect shapes in iris feature images, such as straight lines and circles, and for iris analysis, it can detect the inner and outer circles of the iris. The inner circle diameter here corresponds to the diameter of the pupil, which is a key parameter reflecting the size of the pupil; the outer circle diameter corresponds to the diameter of the entire iris and can be regarded as the boundary of the iris.

[0132] The inner and outer circle diameters obtained by Hough transform can be further used to calculate the ratio of the inner circle diameter to the outer circle diameter, which can be used as a measure of the degree of pupil scaling.

[0133] S26. Using a correction function, taking the radius ratio of the inner and outer edges of the iris as a parameter, radial nonlinear sampling is performed to correct the uneven scaling of the pupil.

[0134] Since the camera of the mobile phone is constantly recording, an iris feature image stream is generated. The image is extracted from the iris feature image stream at a certain step length, and then the pupil radius and iris radius in the iris feature image are obtained by Hough transform, and the ratio of the two is calculated, and then a point in the iris is selected as the reference point. The reference point corresponds to a normalized radius in each extracted image. It should be noted here that the normalized radius refers to the radius corresponding to the reference point in the iris area in the dual dimensionless polar coordinate system. When the degree of pupil scaling is different, the normalized radius corresponding to the same reference point in different iris feature images is different.

[0135] The ratios of the pupil radius to the iris radius of each extracted image are sorted, and the image with the smallest ratio is selected, which is the image with the pupil contracted to the minimum and the iris texture area largest.

[0136] Then the correction function F(S i ), r i =F(S i ),

[0137] Among them, r is the normalized radius, S i is the ratio of pupil radius to iris radius, a j is the coefficient. As the value of j increases, the order of terms increases and the accuracy increases. Due to the influence of mobile phone performance and the fact that the accuracy requirement is not too high, the third term can be used. Therefore, F(S i ).

[0138] S3. Input the iris feature image and the iris center into the convolutional neural network recognition model to obtain the radial pattern information and the radial pattern intersection position information of the iris feature image, wherein the radial pattern information includes position information and angle information, and the convolutional neural network can use the iris center as the convolution starting point and perform convolution along the radial lines from the iris center to the outside.

[0139] Specifically, refer to Figure 6 In one embodiment, S3 includes the following sub-steps:

[0140] S31. Acquire the iris center position based on the corrected iris feature image, and establish a polar coordinate system based on the iris center position.

[0141] The edge of the iris is found by an edge detection algorithm (such as Canny edge detection), and then the center of the iris is found by a circle fitting algorithm. Specifically, the edge points can be fitted to a circle model by the least squares method to determine the center coordinates of the iris (x c ,y c). Then a polar coordinate system is established on the iris feature image. In this polar coordinate system, the iris center is used as the origin, and the angle and distance represent the direction of the point and the distance from the origin, respectively. Each pixel in the image can be converted into polar coordinate form, where the angle ranges from 0 to 360 degrees and the radius is from the iris center to the iris edge.

[0142] For example, for any point P(x c ,y c ), whose polar coordinates with the iris center can be calculated by the following equation:

[0143]

[0144] θ=arctan2(yy c ,xx c )

[0145] S32. Based on the iris center position, a plurality of radial sampling paths with equal angular spacing are sequentially arranged in a circumferential direction around the iris center, and based on the iris center position, a circumferential sampling path is arranged outwardly along the radial direction of the radial sampling path.

[0146] In the polar coordinate system of the iris, a number of radial paths with equal angular distances are set along the annular direction around the center of the iris. For example, a radial path can be set every 0.02 degrees, from 0 degrees to 360 degrees, with a total of 18,000 radial paths. Of course, the sampling angular distance and annular distance should be determined according to the pixels of the obtained iris feature image. The center of the image may be oversampled, so virtual pixels can be generated by interpolation.

[0147] Specifically, each radial path can be extended from the center of the iris by fixing the angle θ and then changing the value of the radius r. In this way, the radial path can represent a series of points on the iris, and the polar coordinates of each point are (r, θ), where θ is fixed.

[0148] Then, it is necessary to set annular sampling paths outward along the radial direction of the radial sampling paths. These annular sampling paths can be formed by fixing the radius r and then changing the value of the angle θ.

[0149] Assuming that we select 18,000 annular sampling paths on each radial path, we can find 18,000*18,000 intersection points on the iris. These intersection points will be used for sampling in step S33. Through the above sub-steps, an accurate sampling grid can be formed on the iris feature image.

[0150] S33. Sampling the iris feature image based on the intersection points of each annular sampling path and radial sampling path.

[0151] The coordinates of the intersection of the annular sampling path and the radial sampling path can be directly expressed by polar coordinates, and the coordinates of each intersection are (r, θ), where r is the radius and θ is the angle. At the location of each intersection, the corresponding pixel value or pixel intensity can be obtained from the iris feature image. This process can be performed by an interpolation method to accurately obtain the pixel information of each intersection on the iris feature image. The information of each sampling point will be stored for subsequent use. This information includes but is not limited to pixel intensity, color, texture features, etc. The information of these sampling points will be used to construct a new feature image in subsequent steps.

[0152] S34. Rearrange the obtained sampling point information into a new feature image, wherein each row of pixel points of the new feature image corresponds to each sampling point on the same radial sampling path on the original iris feature image and has the same order, and each column of pixel points corresponds to each sampling point on the same circumferential sampling path on the original iris feature image and has the same order.

[0153] For example, we first need to initialize an empty feature image matrix. If there are 18,000 radial sampling paths and 18,000 annular sampling paths on each radial path, the size of the new feature image will be 18,000 × 18,000.

[0154] Each intersection point information sampled from the iris feature image will be arranged as a row of the new feature image according to its order on the radial path. For example, all sampling points corresponding to the intersection points on the first radial path will become the first row of the new feature image.

[0155] The sampling point information corresponding to the sampling points on each annular path will be arranged in order as columns of the new feature image. For example, all sampling points corresponding to the sampling points on the first annular sampling path will become the first column of the new feature image.

[0156] Then, according to these arrangement rules, the new feature image is filled using the sampling point information on the iris feature image. The filling can be completed using information such as pixel intensity, color, and texture features of the sampling points.

[0157] S35. Input the new feature image into the convolutional neural network and perform a convolution operation for filtering.

[0158] Here you can select one or more convolution kernels, depending on the iris features you want to detect, such as texture or intersections. Then define the convolution layer parameters, which include determining the size, stride, and padding of the convolution kernel. Then perform the convolution operation, and the convolution kernel is applied to the new feature image, filtering by sliding the convolution kernel on the image and performing a dot product operation.

[0159] For example, the convolution kernel is placed in the upper left corner of the new feature image, and each value in the convolution kernel is multiplied by the covered pixel value, and then these products are added. This operation is repeated on the entire image, and the convolution kernel moves along the image. The result of the convolution operation will be a new feature map, where each pixel represents a specific feature of the corresponding area in the new feature image.

[0160] S36. Set the activation function and perform the maximum pooling operation.

[0161] Here, the activation function set as ReLU activation function is taken as an example. This function can help strengthen features and increase nonlinearity, so that the model can capture more complex feature relationships. ReLU can return all negative values ​​to zero without affecting positive values, increasing the nonlinear ability of the model and helping to highlight iris features.

[0162] The specific implementation process is as follows:

[0163] def relu(x):

[0164] return max(0,x)

[0165] activated_feature_map=[[relu(pixel)for pixel in row]for row inconvoluted_feature_map]

[0166] Next, a maximum pooling operation is performed to reduce the dimension of the feature while retaining the main features. This step usually uses a 2x2 window size and selects the maximum value in the window as the representative.

[0167] The specific implementation process is as follows:

[0168]

[0169] Through ReLU activation and max pooling operations, the feature image is further refined and ready to be input into the next convolutional layer for more advanced feature extraction.

[0170] S37. Select a convolution kernel for detecting horizontal lines and a convolution kernel for detecting intersections and perform convolution operations on the pooled images to obtain spacing features and intersection features.

[0171] First, select the convolution kernel for detecting horizontal lines. In iris feature image analysis, it is necessary to detect certain specific horizontal line structures, which helps to analyze iris texture. A common convolution kernel for detecting horizontal lines can be shown as follows:

[0172]

[0173] Then, a convolution kernel is selected for detecting intersections, which is also an important step in iris analysis because these intersections may be associated with iris features. A common convolution kernel for detecting intersections can be shown as follows:

[0174]

[0175] Use the convolution kernel defined above to convolve the pooled image obtained from step S36. This can be done using a convolution function. For example:

[0176] def convolve(image,kernel):

[0177] #Implementation of convolution operation

[0178] #Return the convolved image ...

[0180] horizontal_feature=convolve(pooled_feature_map,horizontal_kernel)

[0181] intersection_feature=convolve(pooled_feature_map, intersection_kernel).

[0182] Through these two convolution operations, the horizontal line features and intersection features in the iris feature image are obtained.

[0183] S38. Perform an inverse transformation from the rectangular coordinate system to the polar coordinate system based on the spacing features and the intersection features to obtain the radial pattern information and the radial pattern intersection position information in the iris feature image.

[0184] First, determine the inverse transformation function from the rectangular coordinate system to the polar coordinate system. Taking into account the iris center position, we can define the inverse transformation function to convert the spacing features and intersection features back to the polar coordinate system. This can be achieved by the following mathematical expression:

[0185]

[0186] θ=arctan2(y,x)

[0187] Where r and θ are the radius and angle in the polar coordinate system, respectively, and x and y are the coordinates in the rectangular coordinate system.

[0188] Using the above inverse transformation, the spacing features obtained in S37 can be mapped back to the polar coordinate system. For example:

[0189] def inverse_transform(x,y):

[0190] r = np.sqrt(x**2+y**2)

[0191] theta = np.arctan2(y,x)

[0192] return r,theta

[0193] radial_positions=[inverse_transform(x,y)for x,y in spacing_features]

[0194] Similarly, the intersection feature can be mapped back to the polar coordinate system:

[0195] intersection_positions=[inverse_transform(x,y)for x,y inintersection_features]

[0196] Through the above inverse transformation, the polar coordinate position corresponding to the rectangular coordinate system in the iris feature image can be obtained.

[0197] S4. Match the radial pattern information and the intersection position with the pre-stored information in the database to obtain the corresponding identity.

[0198] Specifically, refer to Figure 7 In one embodiment, S4 includes the following sub-steps:

[0199] S41. Loading pre-stored information from a database, wherein the pre-stored information database includes a series of iris radial pattern information and intersection positions of known identities.

[0200] The database contains a series of iris radial pattern information and intersection locations for known identities. Each entry is associated with a specific person's identity, including the radial pattern image of their iris, the location of the intersections, etc.

[0201] S42. Compare the newly acquired iris feature information with the iris features in the database, and calculate the similarity or distance.

[0202] Compare with each pre-stored iris feature in the database. Different comparison algorithms, such as Euclidean distance, cosine similarity, etc., can be used to quantify the similarity between the newly acquired iris and the pre-stored iris. The above steps will result in an array of distances or similarities, each element of which represents the similarity between the newly acquired iris feature and a pre-stored feature. If all distances or similarities do not meet the predetermined threshold, the system may trigger an exception handling process, such as requesting additional identity verification.

[0203] S43. Based on the comparison result, the iris feature in the database that is most similar to or closest to the new iris feature information is matched, and the corresponding identity tag is retrieved.

[0204] According to the similarity or distance calculated in step S42, the system selects the most similar or closest iris feature. Based on the matched iris feature, the system further retrieves the identity tag associated with the feature in the database.

[0205] S44. Combine the retrieved identity tag with other authentication information to improve the security of identification.

[0206] For example, suppose that in addition to iris recognition, there are two additional authentications: fingerprint recognition and password verification. Multiple authentication information is combined to enhance the security of identification. This not only increases the difficulty for attackers to deceive the system, but also makes the system's identification of legitimate users more accurate and reliable.

[0207] S45. After confirming the identity, perform corresponding operations according to the application scenario.

[0208] Once the identity is confirmed, the smartphone automatically unlocks, allowing the user to access their private data and apps.

[0209] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0210] In one embodiment, a computer device is provided, which may be a server. The computer device includes a processor, a memory, a network interface, and a database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used for data related to the steps and methods executed by the processor in the above-mentioned mobile phone. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the method executed by the processor in the mobile phone in the above-mentioned embodiment is implemented.

[0211] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method executed by the processor in the mobile phone in the above embodiment is implemented.

[0212] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method executed by the processor in the mobile phone of the above embodiment is implemented, for example Figure 1 Steps S1 to S4 are shown.

[0213] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments of the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0214] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0215] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A mobile phone based on biometrics, characterized in that: The invention comprises a mobile phone body, a camera device and a processor, wherein the camera device is installed on the mobile phone body and is used to obtain image information from the outside world, and the processor can perform the following steps: S1. Acquire image information and preprocess the image information to obtain an iris feature image; S2. performing standardization correction on the iris feature image and obtaining the iris center; S3. Inputting the iris feature image and the iris center into the convolutional neural network recognition model to obtain the radial pattern information and the radial pattern intersection position information of the iris feature image, wherein the radial pattern information includes the position information and the angle information, and the convolutional neural network can take the iris center as the convolution starting point and perform convolution along the radial lines outward from the iris center; S4. Matching the radial pattern information and the intersection position with the pre-stored information in the database to obtain the corresponding identity; The S3 comprises the following steps: S31. Obtaining the iris center position based on the corrected iris feature image, and establishing a polar coordinate system based on the iris center position; S32. Based on the iris center position, a plurality of radial sampling paths with equal angular spacing are sequentially set in a circular direction around the iris center, and a circular sampling path is set outward along the radial direction of the radial sampling path based on the iris center position; S33. Sampling the iris feature image based on the intersection of each annular sampling path and the radial sampling path; S34. Rearrange the obtained sampling point information into a new feature image, wherein each row of pixel points of the new feature image corresponds to each sampling point on the same radial sampling path on the original iris feature image and has the same order, and each column of pixel points corresponds to each sampling point on the same circumferential sampling path on the original iris feature image and has the same order; S35. Input the new feature image into the convolutional neural network and perform a convolution operation for filtering; S36. Set the activation function and perform the maximum pooling operation; S37. Select a convolution kernel for detecting horizontal lines and a convolution kernel for detecting intersections and perform convolution operations on the pooled images to obtain spacing features and intersection features; S38. Perform an inverse transformation from the rectangular coordinate system to the polar coordinate system based on the spacing features and the intersection features to obtain the radial pattern information and the radial pattern intersection position information in the iris image.

2. The mobile phone based on biometrics according to claim 1, characterized in that: The S2 comprises the following sub-steps: S21. Perform instance segmentation on the iris feature image to obtain a preliminary screening instance of the iris; S22. Extracting iris edge information from the iris feature image, locating the center of the inner and outer iris circles, and calculating geometric parameters; wherein the geometric parameters are the radius, or the major axis and the minor axis; S23. Calculate the center coordinate deviation of the inner circle and the outer circle, and then determine the physiological deviation or sampling deviation based on the deviation; S24. Standardizing the iris feature image based on deviation type and geometric shape classification; S25. Calculate the ratio of the inner and outer iris radii as a measure of the degree of pupil scaling; S26. Using a correction function, taking the radius ratio of the inner and outer edges of the iris as a parameter, radial nonlinear sampling is performed to correct the uneven scaling of the pupil.

3. The biometric mobile phone according to claim 2, characterized in that: The S23 comprises the following sub-steps: S231. Calculate the center coordinate deviation of the inner circle and the outer circle; S232. Determine the type of deviation; S233. Analyze and classify the causes of deviations; S234. Adjust iris positioning according to the deviation; S235. Store the adjusted iris information.

4. The biometric mobile phone according to claim 3, characterized in that: The S24 comprises the following sub-steps: S241. If the iris feature image is an ellipse, obtain the geometric parameters of the ellipse iris and calculate the transformation ratio from the ellipse to the circle; S242. Mapping the ellipse to the circle based on the affine transformation, and interpolating the transformed iris feature image based on the transformation ratio; S243. Calibrate the transformed iris feature image to ensure accuracy.

5. The biometric mobile phone according to claim 1, characterized in that: The S4 comprises the following steps: S41. Loading pre-stored information from a database, wherein the pre-stored information database includes a series of iris radial pattern information and intersection locations of known identities; S42. Compare the newly acquired iris feature information with the iris features in the database and calculate the similarity or distance; S43. Based on the comparison result, the iris feature in the database that is most similar to or closest to the new iris feature information is matched, and the corresponding identity tag is retrieved; S44. combining the retrieved identity tag with other identity verification information to improve the security of identification; S45. After confirming the identity, perform corresponding operations according to the application scenario.

6. A computer device, characterized in that: It includes: one or more processors; Memory; One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to: execute the steps performed by the processor of the mobile phone according to any one of claims 1 to 5.

7. A computer-readable storage medium, characterized in that: The storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set are loaded and executed by the processor to implement: the steps performed by the processor of the mobile phone as described in any one of claims 1 to 5.

Images