Iris recognition method and device

Abstract

The present application relates to the field of artificial intelligence technology, and specifically discloses an iris recognition method and device, wherein the method comprises: obtaining standardized three-dimensional iris image data of an iris to be recognized; using a two-dimensional convolutional neural network to extract and classify features of the standardized three-dimensional iris image data, and obtaining target feature data and a target user label corresponding to the iris to be recognized, wherein the two-dimensional convolutional neural network comprises a three-dimensional image processing layer, and the three-dimensional image processing layer performs dimensionality reduction processing on the standardized three-dimensional iris image data to obtain multiple two-dimensional iris image data, and the multiple two-dimensional iris image data include oblique section data; retrieving iris feature data corresponding to the target user label from a three-dimensional iris image database, wherein the three-dimensional iris image database stores iris feature data of each user among multiple users; generating an iris recognition result according to the target feature data and the retrieved iris feature data. The above scheme can improve the accuracy of iris recognition.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to an iris recognition method and device. Background Art

[0002] At present, iris recognition technology solutions mainly focus on the processing of two-dimensional digital images, for example, using Local Binary Pattern (LBP) algorithm, Support Vector Machine (SVM) algorithm, filter, two-dimensional Convolutional Neural Networks (CNN) algorithm and various improved algorithms.

[0003] Regardless of the algorithm used or how it is optimized, the core features required by the algorithm are extracted based on the two-dimensional image of the iris. The two-dimensional convolutional neural network is the most widely used algorithm in existing two-dimensional image processing methods. For a certain type of feature, a 2×2 two-dimensional convolution kernel can only extract data features in four directions, and a 3×3 two-dimensional convolution kernel can extract data features in eight directions at most, and the extracted feature data is relatively small. Therefore, the recognition accuracy of iris recognition technology based on two-dimensional image processing is not high enough.

[0004] To address the above problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of the present application provide an iris recognition method and device to solve the problem that the recognition accuracy of the iris recognition technology based on two-dimensional image processing in the prior art is not high enough.

[0006] The embodiment of the present application provides an iris recognition method, including: obtaining standardized three-dimensional iris image data of an iris to be recognized; using a two-dimensional convolutional neural network to perform feature extraction and classification on the standardized three-dimensional iris image data to obtain target feature data and a target user label corresponding to the iris to be recognized, wherein the two-dimensional convolutional neural network includes a three-dimensional image processing layer, and the three-dimensional image processing layer is used to perform dimensionality reduction processing on the standardized three-dimensional iris image data to obtain multiple two-dimensional iris image data, and the multiple two-dimensional iris image data include oblique section data; retrieving iris feature data corresponding to the target user label from a three-dimensional iris image database, wherein the three-dimensional iris image database stores iris feature data of each user among multiple users; and generating an iris recognition result according to the target feature data and the retrieved iris feature data.

[0007] In one embodiment, obtaining standardized three-dimensional iris image data of an iris to be identified includes: receiving original ultrasonic data of the iris to be identified; digitizing the original ultrasonic data; and generating standardized three-dimensional iris image data based on the digitized ultrasonic data, wherein the digitized ultrasonic data forms a curved surface in a three-dimensional coordinate system, and the standardized three-dimensional iris image data forms a rectangular parallelepiped in the three-dimensional coordinate system, the rectangular parallelepiped includes the curved surface, and the value of the part of the rectangular parallelepiped except the curved surface is zero.

[0008] In one embodiment, the size of the three-dimensional image processing kernel in the three-dimensional image processing layer is 3×3×3; accordingly, the three-dimensional image processing layer is used to perform dimensionality reduction processing on the standardized three-dimensional iris image data to obtain multiple two-dimensional iris image data, including: using the three-dimensional image processing layer to extract multiple three-dimensional iris image data with a size of 3×3×3 from the standardized three-dimensional iris image data; for each three-dimensional iris image data in the multiple three-dimensional iris image data, vertically cross-sectioning is performed along the length, width and height directions of each three-dimensional iris image data to extract three cross-section data with a size of 3×3, and eight vertex data of each three-dimensional iris image data that have not been extracted are combined with the center point data into oblique cross-section data with a size of 3×3 to obtain multiple two-dimensional iris image data corresponding to the standardized three-dimensional iris image data.

[0009] In one embodiment, the three-dimensional iris image database further stores standardized three-dimensional iris image data of each user among a plurality of users and standardized three-dimensional iris image data of each non-user among a plurality of non-users; the two-dimensional convolutional neural network is obtained by training based on data in the three-dimensional iris image database; the standardized three-dimensional iris image data is extracted and classified by using the two-dimensional convolutional neural network to obtain target feature data and target user labels corresponding to the iris to be identified, including: inputting the standardized three-dimensional iris image data into the trained two-dimensional convolutional neural network to obtain the target feature data, category and identification corresponding to the iris to be identified, wherein the category is user or non-user, and correspondingly, the identification is a user label or a non-user number; when the category corresponding to the iris to be identified is non-user, generating recognition failure data, wherein the recognition failure data is used to indicate that the identity of the iris to be identified cannot be identified.

[0010] In one embodiment, an iris recognition result is generated based on the target feature data and the retrieved iris feature data, including: calculating the similarity between the target feature data and the retrieved iris feature data; when the similarity between the target feature data and the retrieved iris feature data is greater than a preset similarity, determining that the match is successful and outputting a target user tag.

[0011] In one embodiment, calculating the similarity between the target feature data and the retrieved iris feature data includes: calculating the similarity between the target feature data and the retrieved iris feature data according to the following formula:

[0012]

[0013] Among them, cos(M1,M2) is the cosine similarity between the target feature data and the retrieved iris feature data, M1 is the target feature data, M2 is the retrieved iris feature data, M 1i is the element in the target feature data, M 2i is the element in the retrieved iris feature data, and n is the number of elements in the feature data.

[0014] The embodiment of the present application also provides an iris recognition device, including: an acquisition module, used to acquire standardized three-dimensional iris image data of an iris to be recognized; an extraction and classification module, used to perform feature extraction and classification on the standardized three-dimensional iris image data using a two-dimensional convolutional neural network, and obtain target feature data and a target user label corresponding to the iris to be recognized, wherein the two-dimensional convolutional neural network includes a three-dimensional image processing layer, and the three-dimensional image processing layer is used to perform dimensionality reduction processing on the standardized three-dimensional iris image data to obtain multiple two-dimensional iris image data, and the multiple two-dimensional iris image data include oblique section data; a retrieval module, used to retrieve iris feature data corresponding to the target user label from a three-dimensional iris image database, wherein the three-dimensional iris image database stores iris feature data of each user among multiple users; and a generation module, used to generate an iris recognition result according to the target feature data and the retrieved iris feature data.

[0015] An embodiment of the present application further provides a computer device, including a processor and a memory for storing instructions executable by the processor, wherein the processor implements the steps of the iris recognition method described in any of the above embodiments when executing the instructions.

[0016] An embodiment of the present application further provides a computer-readable storage medium having computer instructions stored thereon, and when the instructions are executed, the steps of the iris recognition method described in any of the above embodiments are implemented.

[0017] In an embodiment of the present application, an iris recognition method is provided, which can obtain standardized three-dimensional iris image data of an iris to be recognized, and input the standardized three-dimensional iris image data into a two-dimensional convolutional neural network. The two-dimensional convolutional neural network includes a three-dimensional image processing layer. The three-dimensional image processing layer can perform dimensionality reduction processing on the standardized three-dimensional iris image data to obtain multiple two-dimensional iris image data. The multiple two-dimensional iris image data include oblique section data. The convolution layer in the two-dimensional convolutional neural network can use a two-dimensional convolution kernel to perform convolution processing on the multiple two-dimensional iris image data to achieve feature extraction and classification of the standardized three-dimensional iris image data to obtain target feature data and a target user label corresponding to the iris to be recognized. After that, the iris feature data corresponding to the target user label can be retrieved from a three-dimensional iris image database storing the iris feature data of each user among multiple users, and an iris recognition result is generated based on the target feature data and the retrieved iris feature data. In the above scheme, iris recognition is performed based on three-dimensional iris image data, which can improve the accuracy of iris recognition; first, the three-dimensional iris image data is reduced in dimension using a three-dimensional image processing layer to obtain multiple two-dimensional iris image data, and then the multiple two-dimensional iris image data are extracted and classified using a two-dimensional convolutional neural network. Compared with the three-dimensional convolutional neural network, the resources and time required for calculation can be significantly reduced, thereby reducing the recognition cost and improving the recognition efficiency; further, since the multiple two-dimensional iris image data include oblique section data, the oblique section data includes the center point and vertex data of the standardized three-dimensional iris image data or the image data subset in the standardized three-dimensional iris image data, the iris features in the inclined direction can be characterized, thereby further improving the accuracy of iris recognition. The above scheme solves the technical problem of low accuracy of iris recognition based on two-dimensional images, and achieves the technical effect of effectively improving the recognition accuracy and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the present application, constitute a part of the present application, and do not constitute a limitation of the present application. In the drawings:

[0019] Figure 1 A flow chart of an iris recognition method in an embodiment of the present application is shown;

[0020] Figure 2 A schematic diagram showing a method for iris recognition in an embodiment of the present application using a three-dimensional image processing layer to process a three-dimensional image;

[0021] Figure 3 A schematic diagram showing cross-sectional data obtained after a three-dimensional image is processed by a three-dimensional image processing layer in an iris recognition method in an embodiment of the present application;

[0022] Figure 4A schematic diagram showing a flow chart of an iris recognition method in an embodiment of the present application is shown;

[0023] Figure 5 A schematic diagram of a two-dimensional convolutional neural network in an iris recognition method in an embodiment of the present application is shown;

[0024] Figure 6 A schematic diagram of an iris recognition device in an embodiment of the present application is shown;

[0025] Figure 7 A schematic diagram of a computer device in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0026] The principles and spirit of the present application will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present application, and are not intended to limit the scope of the present application in any way. On the contrary, these embodiments are provided to make the disclosure of the present application more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0027] Those skilled in the art know that the embodiments of the present application can be implemented as a system, device, method or computer program product. Therefore, the present application can be implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0028] An embodiment of the present application provides an iris recognition method. Figure 1 A flow chart of an iris recognition method in an embodiment of the present application is shown. Although the present application provides method operation steps or device structures as shown in the following embodiments or drawings, more or fewer operation steps or module units may be included in the method or device based on routine or without creative labor. In the steps or structures where there is no necessary causal relationship logically, the execution order of these steps or the module structure of the device is not limited to the execution order or module structure described in the embodiments of the present application and shown in the drawings. When the method or module structure is applied to an actual device or terminal product, it can be connected according to the method or module structure shown in the embodiments or drawings for sequential execution or parallel execution (for example, a parallel processor or a multi-threaded processing environment, or even a distributed processing environment).

[0029] Specifically, Figure 1 As shown, an iris recognition method provided by an embodiment of the present application may include the following steps.

[0030] Step S101, obtaining standardized three-dimensional iris image data of an iris to be recognized.

[0031] Iris recognition refers to a biometric identification technology that identifies the annular area between the black pupil and the white sclera of the human eye. It is an identity authentication method that can use the rich texture information of the human iris to obtain identity information. Its core technology lies in extracting texture features with high discriminability.

[0032] The method in the embodiment of the present application can be applied to a server. The server can directly receive the standardized three-dimensional iris image data of the iris to be identified from the client, or receive the three-dimensional iris image data to be identified from the client, and perform standardization processing on the received three-dimensional iris image data to obtain the standardized three-dimensional iris image data. In the standardized three-dimensional iris image data, each pixel can be represented by a unit cube, and the coordinates of the unit cube in the three-dimensional coordinate system are the position of the pixel in the three-dimensional space. The standardized three-dimensional iris image data can form a rectangular parallelepiped area in the three-dimensional coordinate system to facilitate processing by the convolutional neural network.

[0033] Step S102, using a two-dimensional convolutional neural network to perform feature extraction and classification on the standardized three-dimensional iris image data, to obtain target feature data and a target user label corresponding to the iris to be identified.

[0034] Specifically, the standardized three-dimensional iris image data of the iris to be identified can be input into a trained two-dimensional convolutional neural network for feature extraction and classification, thereby obtaining target feature data and target user labels corresponding to the iris to be identified. The two-dimensional convolutional neural network may include a three-dimensional image processing layer. The three-dimensional image processing layer may be used to perform dimensionality reduction processing on the standardized three-dimensional iris image data to obtain multiple two-dimensional iris image data. The multiple two-dimensional iris image data include oblique section data. The oblique section data may include data on the center point and vertices of the standardized three-dimensional iris image data, or may include data on the center point and vertices of a subset of image data in the standardized three-dimensional iris image data, thereby characterizing iris features in an oblique direction. The two-dimensional convolutional neural network may include one or more convolutional layers, and the convolution kernels of each convolutional layer are all two-dimensional convolutional kernels. The multiple two-dimensional iris image data output by the three-dimensional image processing layer may be convoluted to achieve feature extraction and classification, thereby obtaining target feature data and target user labels corresponding to the iris to be identified.

[0035] Step S103: retrieve iris feature data corresponding to the target user tag from the three-dimensional iris image database.

[0036] Step S104, generating an iris recognition result according to the target feature data and the retrieved iris feature data.

[0037] The 3D iris image database may store iris feature data of each of the multiple users. For example, the 3D iris image database may store user tags, iris feature data, and the corresponding relationship between the user tags and the iris feature data of each of the multiple users. After obtaining the target user tag, the iris feature data corresponding to the target user tag may be retrieved from the 3D iris image database.

[0038] After the iris feature data corresponding to the target user tag is retrieved, an iris recognition result can be generated based on the target feature data and the retrieved iris feature data. For example, the target feature data and the retrieved iris feature data can be matched, and if the match is successful, a prompt message indicating the successful match and the target user tag are output.

[0039] In this embodiment, iris recognition is used as an example to describe an iris recognition method based on a three-dimensional image. Those skilled in the art can understand that the method in the above embodiment can also be applied to other biometric recognitions, including but not limited to: fingerprint recognition, face recognition, and palm print recognition.

[0040] The method in the above embodiment performs iris recognition based on three-dimensional iris image data, which can improve the accuracy of iris recognition. First, the three-dimensional iris image data is reduced in dimension using a three-dimensional image processing layer to obtain multiple two-dimensional iris image data, and then a two-dimensional convolutional neural network is used to extract and classify features of the obtained multiple two-dimensional iris image data. Compared with the three-dimensional convolutional neural network, the resources and time required for calculation can be significantly reduced, thereby reducing the recognition cost and improving the recognition efficiency. Furthermore, since the multiple two-dimensional iris image data include oblique section data, the oblique section data includes the standardized three-dimensional iris image data or the center point and vertex data of the image data subset in the standardized three-dimensional iris image data, the iris features in the inclined direction can be characterized, thereby further improving the accuracy of iris recognition.

[0041] In some embodiments of the present application, obtaining standardized three-dimensional iris image data of an iris to be identified may include: receiving original ultrasonic data of the iris to be identified; digitizing the original ultrasonic data; and generating standardized three-dimensional iris image data based on the digitized ultrasonic data, wherein the digitized ultrasonic data forms a curved surface in a three-dimensional coordinate system, and the standardized three-dimensional iris image data forms a rectangular parallelepiped in the three-dimensional coordinate system, the rectangular parallelepiped includes the curved surface, and the value of the part of the rectangular parallelepiped except the curved surface is zero.

[0042] Specifically, the server can receive the original ultrasonic data of the iris to be identified. The original ultrasonic data can be a three-dimensional grayscale image. The original ultrasonic data can be digitized to obtain the position coordinates and grayscale values ​​corresponding to each pixel point. Afterwards, the digitized ultrasonic data is arranged in a three-dimensional coordinate system to obtain a curved surface similar to a hemisphere. A rectangular parallelepiped can be established to surround the curved surface according to the length, height and width of the curved surface in the three-dimensional coordinate system. The length, height and width of the rectangular parallelepiped are equal to the length, height and width of the curved surface, and the numerical value of the part of the rectangular parallelepiped except the curved surface is set to 0 to obtain standardized three-dimensional iris image data. In the above manner, the three-dimensional grayscale image can be standardized to obtain standardized three-dimensional iris image data, which is convenient for convolutional neural network to process.

[0043] In some embodiments of the present application, the size of the three-dimensional image processing kernel in the three-dimensional image processing layer is 3×3×3; accordingly, using the three-dimensional image processing layer to perform dimensionality reduction processing on the standardized three-dimensional iris image data to obtain multiple two-dimensional iris image data can include: using the three-dimensional image processing layer to extract multiple stereoscopic iris image data with a size of 3×3×3 from the standardized three-dimensional iris image data; for each stereoscopic iris image data in the multiple stereoscopic iris image data, vertically sectioning along the length, width and height directions of each stereoscopic iris image data to extract three section data with a size of 3×3, and combining the eight vertex data of each stereoscopic iris image data that has not been extracted with the center point data into oblique section data with a size of 3×3, to obtain multiple two-dimensional iris image data corresponding to the standardized three-dimensional iris image data.

[0044] Specifically, the three-dimensional image processing layer may include a three-dimensional image processing core. The size of the three-dimensional image processing core may be 3×3×3, and the unit of the size may be a unit cube in a three-dimensional coordinate system, corresponding to one pixel, so the three-dimensional image processing core corresponds to 27 pixels. The three-dimensional image processing layer may be used to extract multiple two-dimensional iris image data from the standardized three-dimensional iris image data. For example, the three-dimensional image data processing core may be used to move along the length, width, and height directions of the cuboid corresponding to the standardized three-dimensional iris image data at a preset step length until all the pixels of the standardized three-dimensional iris image data are traversed. Each time it moves, a 3×3×3 stereoscopic iris image data may be obtained. For example, for the standardized three-dimensional iris image data of 3×3×4, it is only necessary to move the three-dimensional image processing core along the Z-axis direction to obtain two 3×3×3 stereoscopic iris image data.

[0045] For each obtained 3D iris image data, the 3D image processing core performs cross-sectioning. Vertical cross-sectioning can be performed along the length, width and height directions of each cubic iris image data to extract three cross-section data of size 3×3, which are parallel to the XY plane, XZ plane and YZ plane respectively. Please refer to Figure 2 , shows a schematic diagram of processing a three-dimensional image using a three-dimensional image processing layer. Figure 2 , a schematic diagram of vertical cross-sectioning of the 3D iris image data along the length, width and height directions is shown. Figure 2 As shown, after the vertical section is made along the length, width and height directions, the pixel data at the eight vertices of the cube are not intercepted. Therefore, the pixel data of the eight vertices that are not intercepted ( Figure 2 The dotted pixel cube in the figure) and the center pixel data of the cube ( Figure 2 The bold solid cube in the figure constitutes the oblique section data with a size of 3×3.

[0046] Please refer to Figure 3 , shows a schematic diagram of cross-sectional data obtained after a three-dimensional image is processed by a three-dimensional image processing layer in an iris recognition method in an embodiment of the present application. Figure 3 As shown, the 27 pixels in the cubic iris image data can be numbered. After the cubic iris image data is sectioned, four 3×3 section data can be obtained, namely Figure 3 The cross section, vertical section, longitudinal section and oblique section in the figure. The oblique section is composed of 8 pixels in the oblique direction (i.e., the 8 vertices of the cube) and the central pixel. Figure 3 Only one arrangement of the oblique cross-section is shown. The oblique cross-section in the embodiment of the present application can also be arranged in other ways, as long as the central pixel point 14 of the cube is located at the center of the oblique cross-section, and the 8 vertex data can be randomly arranged around the center point.

[0047] In the above embodiment, after the standardized three-dimensional iris image data is subjected to dimensionality reduction processing by the three-dimensional image processing core in the three-dimensional image processing layer, four cross-sectional data corresponding to each of the multiple cubic iris image data can be obtained, thereby obtaining multiple two-dimensional iris image data corresponding to the standardized three-dimensional iris image data. The multiple two-dimensional iris image data include oblique cross-sectional data, which can characterize the features in the oblique direction, and is conducive to improving the accuracy of iris recognition.

[0048] It will be appreciated by those skilled in the art that other sizes of three-dimensional image processing cores may also be used, such as 4×4×4 or 5×5×5. When the size of the three-dimensional image processing core changes, the sectioning method of the cubic iris image data may also be adaptively changed. For example, for a 4×4×4 three-dimensional image processing core, the 4×4×4 cubic iris image data may be sectioned twice along the X, Y, and Z directions, respectively, to obtain 6 4×4 oblique section data, and the 6 section data may include other pixel data except for the 8 vertices. Then, the 8 vertices and the 8 pixel data located in the middle of the 4×4×4 cubic iris image data are combined into an oblique section to obtain 4×4 oblique section data. Therefore, for a 4×4×4 three-dimensional image processing core, a 4×4×4 cubic iris image data may be obtained for each movement, and the corresponding 7 section data may be obtained.

[0049] In some embodiments of the present application, the three-dimensional iris image database also stores standardized three-dimensional iris image data of each user among a plurality of users and standardized three-dimensional iris image data of each non-user among a plurality of non-users; the two-dimensional convolutional neural network is trained based on the data in the three-dimensional iris image database; the standardized three-dimensional iris image data is extracted and classified by using the two-dimensional convolutional neural network to obtain target feature data and target user labels corresponding to the iris to be identified, which may include: inputting the standardized three-dimensional iris image data into the trained two-dimensional convolutional neural network to obtain the target feature data, category and identification corresponding to the iris to be identified, wherein the category is user or non-user, and correspondingly, the identification is a user label or a non-user number; in the case where the category corresponding to the iris to be identified is non-user, generating recognition failure data, wherein the recognition failure data is used to indicate that the identity of the iris to be identified cannot be identified.

[0050] Specifically, the three-dimensional iris image database may also store standardized three-dimensional iris image data of each user among multiple users and standardized three-dimensional iris image data of each non-user among multiple non-users. The definition of users and non-users may be adaptively changed according to the application scenario. For example, users may refer to VIP members of a bank or other enterprise, and non-users may refer to ordinary members of a bank or other enterprise. The three-dimensional iris image database may store standardized three-dimensional iris image data of each user among multiple users and standardized three-dimensional iris image data of each non-user among multiple non-users. For example, the three-dimensional iris image database may store standardized three-dimensional iris image data, classifications and identifications of users and non-users. The classifications may include users and non-users. For users, the identification is a user label, and the user label may be an ID number or other unique identification information. For non-users, the identification may be a number. The standardized three-dimensional iris image data, classifications and identifications of users and non-users stored in the three-dimensional iris image database may be used to train the convolutional neural network. For example, the data in the database may be divided into 70% training data sets and 30% test data sets. The training data set and the test data set may adopt a crossover mode, exchanging the training and test data each time the model is retrained to prevent the model from overfitting. Each iris feature data has a unique user / non-user classification and a unique user name label or user number. After the two-dimensional convolutional neural network model is constructed, it is trained using the three-dimensional iris image database. After reaching a certain recognition accuracy, the training is stopped to obtain a trained two-dimensional convolutional neural network. The iris feature data of each user obtained during the training process can also be stored in the three-dimensional iris image database.

[0051] After obtaining the trained two-dimensional convolutional neural network model, the standardized three-dimensional iris image data can be input into the trained two-dimensional convolutional neural network model to obtain the target feature data, category and identification corresponding to the iris to be identified. In the case where the category corresponding to the iris to be identified is non-user, recognition failure data is generated, wherein the recognition failure data is used to indicate that the identity of the iris to be identified cannot be identified. That is, for non-user data, the match fails, or the permission is not granted to the non-user. In the above manner, a large amount of data can be used to train the two-dimensional convolutional neural network, and the trained neural network can be used to extract and classify the standardized three-dimensional iris feature image.

[0052] In some embodiments of the present application, generating an iris recognition result based on the target feature data and the retrieved iris feature data may include: calculating the similarity between the target feature data and the retrieved iris feature data; when the similarity between the target feature data and the retrieved iris feature data is greater than a preset similarity, determining that the match is successful and outputting a target user tag.

[0053] Specifically, after obtaining the target feature data corresponding to the iris to be identified and the retrieved iris feature data corresponding to the target user tag, the similarity between the target feature data and the retrieved iris feature data can be calculated. When the similarity between the target feature data and the retrieved iris feature data is greater than the preset similarity, it is determined that the match is successful and the target user tag is input. In the above manner, the user identified by the convolutional neural network can be further verified, thereby further improving the accuracy of iris recognition.

[0054] In some embodiments of the present application, calculating the similarity between the target feature data and the retrieved iris feature data may include: calculating the similarity between the target feature data and the retrieved iris feature data according to the following formula:

[0055]

[0056] Among them, cos(M1,M2) is the cosine similarity between the target feature data and the retrieved iris feature data, M1 is the target feature data, M2 is the retrieved iris feature data, M 1i is the element in the target feature data, M 2i is the element in the retrieved iris feature data, and n is the number of elements in the feature data.

[0057] In the above embodiment, cosine similarity is used to calculate the similarity between feature vectors. It can be understood by those skilled in the art that other methods in the prior art can also be used to calculate the similarity, and this application does not limit this.

[0058] The above method is described below in conjunction with a specific embodiment. However, it should be noted that this specific embodiment is only for better illustrating the present application and does not constitute an improper limitation on the present application.

[0059] Please refer to Figure 4 , shows a flow chart of an iris recognition method in an embodiment of the present application. In this embodiment, an iris recognition method based on a three-dimensional image model is proposed, and the method is implemented by an iris recognition system, which is composed of an iris image acquisition module, an algorithm module, and a result display module.

[0060] In this specific embodiment, a convolution algorithm method for processing three-dimensional images is proposed. On the basis of using the horizontal, vertical and vertical cross-sectional images, the 8 pixels not obtained by the three cross-sections are rearranged to form a 3×3 pixel map with the center point pixel. The purpose is to extract not only the image features of the horizontal, vertical and vertical cross-sections, but also the image features of the tilt angle direction. In this way, all pixels can be convolved at the same time.

[0061] The iris image acquisition module can be composed of an iris data collector and a three-dimensional iris image construction model. The iris data collector can be an ultrasonic probe that has the function of transmitting and receiving ultrasonic waves. The three-dimensional iris image construction model digitizes the received original ultrasonic data and arranges them in a three-dimensional coordinate system to form a hemispherical curved surface. A cube is then constructed with the length, width and height of the curved surface, and the blank part except the curved surface is digitized to 0 to obtain standardized three-dimensional iris image data.

[0062] The algorithm module can be composed of a three-dimensional iris image database and a 3D Image Processing Core-Convolutional Neural Networks (3D Image Processing Core-Convolutional Neural Networks, a convolutional neural network model based on a three-dimensional image processing core). The three-dimensional iris image database contains user data and non-user data. The database is divided into 70% training database and 30% test database. The training database and the test database adopt a cross-mode, and the training and test data are exchanged each time the model is retrained to prevent the model from overfitting. Each three-dimensional data has a unique user / non-user classification and a unique user name label. After the 3D IPC-CNN model is built, it is trained using the three-dimensional iris image database. After reaching a certain recognition accuracy, stop training.

[0063] Please refer to Figure 5 , shows a schematic diagram of a 3D IPC-CNN model in an iris recognition method in an embodiment of the present application. Figure 5 As shown in Figure 1, the 3DIPC-CNN model can be composed of an input layer, a hidden layer, and an output layer. The input layer is used to input three-dimensional iris image data; the output layer is used to output the extracted features, classification results, and labels. Figure 5 As shown, the hidden layer can be composed of a 3DIPC layer, two convolutional layers, two pooling layers and a fully connected layer.

[0064] Please refer to Figure 3 , showing a schematic diagram of cross-sectional data obtained after processing a three-dimensional image using a three-dimensional image processing layer in an iris recognition method in an embodiment of the present application. Taking the processing of a 3×3×3 three-dimensional image as an example, each pixel of the 3×3×3 three-dimensional image is first represented by a small cube and labeled. Among them, pixel 14 is the center pixel, and pixels 1, 3, 7, 9, 19, 21, 25, and 27 are pixels in the inclined direction. When 3DIPC processes a 3×3×3 three-dimensional image, four 3×3 pixel images will be obtained, which are represented as a cross section, a vertical section, a longitudinal section, and an oblique section. As Figure 3As shown, the cross section is parallel to the XY plane, the vertical section is parallel to the YZ plane, the longitudinal section is parallel to the XZ axis, and the oblique section is not a plane obtained by intercepting the three-dimensional image with a plane, but a regular recombination of the pixels in 8 inclined directions and the center pixel. When rearranging the 8 pixels in the inclined direction with the center pixel, other arrangements can also be adopted. Using the 3×3×3 three-dimensional image processed by 3DIPC, not only the data of the horizontal, vertical, and longitudinal sections are obtained, but also the data of the inclined direction and the center pixel are extracted. Subsequently, the CNN model is used to analyze the extracted four types of interface data, and all the data features of the center pixel in 26 directions in three-dimensional space are obtained.

[0065] Please continue to refer to Figure 4 The result display module can be composed of an iris feature matching model and a result display interface. The iris feature matching model matches the iris features output by the 3DIPC-CNN model with the user iris features in the three-dimensional iris image database, using the cosine similarity measurement formula, which is as follows:

[0066]

[0067] The matching result is obtained. The result display interface displays the matching result obtained by the iris feature matching model.

[0068] In this specific embodiment, the iris recognition and verification method based on the three-dimensional image model may include the following steps.

[0069] Step 1: Use an iris data collector to collect iris ultrasonic data.

[0070] Step 2: Input the collected iris ultrasonic data into a three-dimensional iris image construction model to generate three-dimensional iris image data.

[0071] Step 3: Input the generated three-dimensional iris image data into the trained 3D-CNN neural network model, extract the iris feature data, and output the determined user name.

[0072] Step 4: retrieve the user's iris feature data from the three-dimensional iris image database, input it into the iris feature matching model together with the iris feature data extracted in step 3, and obtain the matching result by calculating the cosine similarity.

[0073] Step 5: Display the matching results obtained in step 4 on the result display interface.

[0074] In the above specific embodiments, a method for iris recognition based on a new three-dimensional image convolution is proposed, in order to address the fact that existing iris recognition technology still stays in the field of two-dimensional image processing, and to optimize the existing three-dimensional image processing method. In order to address the fact that existing iris recognition technology remains in the field of two-dimensional image processing technology, a three-dimensional iris stereo image model is established using three-dimensional iris image data to extract data features in more directions from the three-dimensional image data set. In order to address the fact that the three-vertical cross-section data acquisition method for existing three-dimensional image data processing cannot obtain the features of the central pixel and the pixel in the oblique direction, a 3D image processing core (3DIPC) is proposed, which arranges 27 pixels in four cross-sections of horizontal, vertical, longitudinal, and oblique extraction, rearranges them, and then processes them to obtain more feature data to improve accuracy.

[0075] Based on the same inventive concept, an iris recognition device is also provided in the embodiments of the present application, as described in the following embodiments. Since the principle of solving the problem by the iris recognition device is similar to that of the iris recognition method, the implementation of the iris recognition device can refer to the implementation of the iris recognition method, and the repeated parts will not be repeated. As used below, the term "unit" or "module" can be a combination of software and / or hardware that implements the predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable. Figure 6 is a structural block diagram of an iris recognition device according to an embodiment of the present application. Figure 6 As shown, it includes: an acquisition module 601, an extraction and classification module 602, a call module 603 and a generation module 604. The structure is described below.

[0076] The acquisition module 601 is used to acquire standardized three-dimensional iris image data of the iris to be recognized.

[0077] The extraction and classification module 602 is used to use a two-dimensional convolutional neural network to perform feature extraction and classification on the standardized three-dimensional iris image data to obtain target feature data and a target user label corresponding to the iris to be identified, wherein the two-dimensional convolutional neural network includes a three-dimensional image processing layer, and the three-dimensional image processing layer is used to perform dimensionality reduction processing on the standardized three-dimensional iris image data to obtain multiple two-dimensional iris image data, and the multiple two-dimensional iris image data include oblique section data.

[0078] The retrieving module 603 is used to retrieve iris feature data corresponding to the target user tag from the three-dimensional iris image database, wherein the three-dimensional iris image database stores iris feature data of each user among a plurality of users.

[0079] The generating module 604 is used to generate an iris recognition result according to the target feature data and the retrieved iris feature data.

[0080] In some embodiments of the present application, the size of the three-dimensional image processing kernel in the three-dimensional image processing layer is 3×3×3; accordingly, the extraction and classification module is specifically used to: use the three-dimensional image processing layer to extract multiple stereoscopic iris image data with a size of 3×3×3 from the standardized three-dimensional iris image data; for each stereoscopic iris image data in the multiple stereoscopic iris image data, perform vertical sectioning along the length, width and height directions of each stereoscopic iris image data, extract three section data with a size of 3×3, and combine the eight vertex data of each stereoscopic iris image data that have not been extracted with the center point data into oblique section data with a size of 3×3, so as to obtain multiple two-dimensional iris image data corresponding to the standardized three-dimensional iris image data.

[0081] From the above description, it can be seen that the embodiment of the present application achieves the following technical effects: iris recognition based on three-dimensional iris image data can improve the accuracy of iris recognition; first, the three-dimensional iris image data is reduced in dimension by using a three-dimensional image processing layer to obtain multiple two-dimensional iris image data, and then the multiple two-dimensional iris image data are extracted and classified by using a two-dimensional convolutional neural network, which can significantly reduce the resources and time required for calculation compared to the three-dimensional convolutional neural network, thereby reducing the recognition cost and improving the recognition efficiency; further, since the multiple two-dimensional iris image data include oblique section data, the oblique section data includes the center point and vertex data of the standardized three-dimensional iris image data or the image data subset in the standardized three-dimensional iris image data, so that the iris features in the inclined direction can be characterized, and thus the accuracy of iris recognition can be further improved. The above scheme solves the technical problem of low accuracy of iris recognition based on two-dimensional images, and achieves the technical effect of effectively improving the recognition accuracy and efficiency.

[0082] The present application also provides a computer device, which can be found in Figure 7 The computer device structure diagram of the iris recognition method provided in the embodiment of the present application is shown, and the computer device may specifically include an input device 71, a processor 72, and a memory 73. The memory 73 is used to store processor executable instructions. When the processor 72 executes the instructions, the steps of the iris recognition method described in any of the above embodiments are implemented.

[0083] In this embodiment, the input device may specifically be one of the main devices for information exchange between the user and the computer system. The input device may include a keyboard, a mouse, a camera, a scanner, a light pen, a handwriting input board, a voice input device, etc.; the input device is used to input the original data and the program for processing these numbers into the computer. The input device can also obtain and receive data transmitted from other modules, units, and devices. The processor can be implemented in any appropriate manner. For example, the processor can take the form of a computer-readable medium, a logic gate, a switch, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), a programmable logic controller, and an embedded microcontroller, etc., such as a microprocessor or a processor and a computer-readable program code (such as software or firmware) that can be executed by the (micro) processor. The memory may specifically be a memory device used to store information in modern information technology. The memory may include multiple levels. In a digital system, anything that can store binary data can be a memory; in an integrated circuit, a circuit with a storage function without a physical form is also called a memory, such as a RAM, a FIFO, etc.; in a system, a storage device with a physical form is also called a memory, such as a memory stick, a TF card, etc.

[0084] In this embodiment, the functions and effects specifically realized by the computer device can be explained in comparison with other embodiments and will not be described in detail here.

[0085] In an embodiment of the present application, a computer storage medium based on an iris recognition method is also provided, wherein the computer storage medium stores computer program instructions, and when the computer program instructions are executed, the steps of the iris recognition method described in any of the above embodiments are implemented.

[0086] In this embodiment, the storage medium includes, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a cache, a hard disk (HDD) or a memory card. The memory may be used to store computer program instructions. The network communication unit may be an interface for network connection communication set in accordance with the standard specified by the communication protocol.

[0087] In this embodiment, the functions and effects specifically implemented by the program instructions stored in the computer storage medium can be explained in comparison with other embodiments and will not be repeated here.

[0088] Obviously, those skilled in the art should understand that the modules or steps of the above-mentioned embodiments of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. In this way, the embodiments of the present application are not limited to any specific combination of hardware and software.

[0089] It should be understood that the above description is for illustration and not for limitation. By reading the above description, many embodiments and many applications beyond the examples provided will be apparent to those skilled in the art. Therefore, the scope of the present application should not be determined with reference to the above description, but should be determined with reference to the aforementioned claims and the full scope of equivalents possessed by these claims.

[0090] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the embodiments of the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An iris recognition method, characterized in that: include: Acquire standardized three-dimensional iris image data of the iris to be identified; wherein each pixel in the standardized three-dimensional iris image data is represented by a unit cube, and the coordinates of the unit cube in the three-dimensional coordinate system are the position of the pixel in the three-dimensional space; Using a two-dimensional convolutional neural network to extract features and classify the standardized three-dimensional iris image data to obtain target feature data and a target user label corresponding to the iris to be identified, wherein the two-dimensional convolutional neural network includes a three-dimensional image processing layer, and the three-dimensional image processing layer is used to perform dimensionality reduction processing on the standardized three-dimensional iris image data to obtain multiple two-dimensional iris image data, and the multiple two-dimensional iris image data include oblique section data; Retrieving iris feature data corresponding to the target user tag from a three-dimensional iris image database, wherein the three-dimensional iris image database stores iris feature data of each user among a plurality of users; Generate an iris recognition result according to the target feature data and the retrieved iris feature data; The three-dimensional image processing layer is used to perform dimension reduction processing on the standardized three-dimensional iris image data to obtain a plurality of two-dimensional iris image data, including: Extracting a plurality of stereoscopic iris image data from the standardized three-dimensional iris image data using a three-dimensional image processing core of a certain size in the three-dimensional image processing layer; For each of the multiple 3D iris image data, vertical sections are respectively made along the length, width and height directions of the each 3D iris image data, three section data of corresponding sizes are extracted, and eight vertex data and center point data of each 3D iris image data that are not extracted are combined into oblique section data of corresponding sizes to obtain multiple 2D iris image data corresponding to the standardized 3D iris image data.

2. The method according to claim 1, characterized in that Obtain standardized three-dimensional iris image data of the iris to be recognized, including: Receiving raw ultrasonic data of an iris to be recognized; digitizing the raw ultrasonic data; Based on the digitized ultrasonic data, standardized three-dimensional iris image data is generated, wherein the digitized ultrasonic data forms a curved surface in a three-dimensional coordinate system, and the standardized three-dimensional iris image data forms a rectangular parallelepiped in the three-dimensional coordinate system, the rectangular parallelepiped includes the curved surface, and the value of the part of the rectangular parallelepiped except the curved surface is zero.

3. The method according to claim 1, characterized in that The size of the 3D image processing core in the 3D image processing layer is 3×3×3; Accordingly, the standardized three-dimensional iris image data is subjected to dimensionality reduction processing by using the three-dimensional image processing layer to obtain a plurality of two-dimensional iris image data, including: Extracting a plurality of stereoscopic iris image data with a size of 3×3×3 from the standardized three-dimensional iris image data using the three-dimensional image processing layer; For each of the multiple three-dimensional iris image data, vertical sections are respectively performed along the length, width and height directions of the each of the three-dimensional iris image data to extract three section data with a size of 3×3, and the eight vertex data and the center point data of each of the three-dimensional iris image data that are not extracted are combined into oblique section data with a size of 3×3 to obtain multiple two-dimensional iris image data corresponding to the standardized three-dimensional iris image data.

4. The method according to claim 1, characterized in that: The three-dimensional iris image database also stores standardized three-dimensional iris image data of each user among a plurality of users and standardized three-dimensional iris image data of each non-user among a plurality of non-users; the two-dimensional convolutional neural network is trained based on the data in the three-dimensional iris image database; The standardized three-dimensional iris image data is subjected to feature extraction and classification using a two-dimensional convolutional neural network to obtain target feature data and a target user label corresponding to the iris to be identified, including: Input the standardized three-dimensional iris image data into a trained two-dimensional convolutional neural network to obtain target feature data, category and identification corresponding to the iris to be identified, wherein the category is user or non-user, and correspondingly, the identification is a user label or a non-user number; In a case where the category corresponding to the iris to be identified is non-user, identification failure data is generated, wherein the identification failure data is used to indicate that the identity of the iris to be identified cannot be identified.

5. The method according to claim 1, characterized in that Generating an iris recognition result according to the target feature data and the retrieved iris feature data, including: Calculating the similarity between the target feature data and the retrieved iris feature data; When the similarity between the target feature data and the retrieved iris feature data is greater than a preset similarity, it is determined that the match is successful and the target user tag is output.

6. The method according to claim 5, characterized in that Calculating the similarity between the target feature data and the retrieved iris feature data includes: The similarity between the target feature data and the retrieved iris feature data is calculated according to the following formula: Wherein, cos(M1, M2) is the cosine similarity between the target feature data and the retrieved iris feature data, M1 is the target feature data, M2 is the retrieved iris feature data, and M 1i is an element in the target feature data, M 2i is an element in the retrieved iris feature data, and n is the number of elements in the feature data.

7. An iris recognition device, characterized in that: include: An acquisition module, used for acquiring standardized three-dimensional iris image data of an iris to be identified; wherein each pixel in the standardized three-dimensional iris image data is represented by a unit cube, and the coordinates of the unit cube in the three-dimensional coordinate system are the position of the pixel in the three-dimensional space; an extraction and classification module, configured to perform feature extraction and classification on the standardized three-dimensional iris image data using a two-dimensional convolutional neural network to obtain target feature data and a target user label corresponding to the iris to be identified, wherein the two-dimensional convolutional neural network includes a three-dimensional image processing layer, and the three-dimensional image processing layer is configured to perform dimensionality reduction processing on the standardized three-dimensional iris image data to obtain a plurality of two-dimensional iris image data, wherein the plurality of two-dimensional iris image data includes oblique section data; A retrieving module, configured to retrieve iris feature data corresponding to the target user tag from a three-dimensional iris image database, wherein the three-dimensional iris image database stores iris feature data of each user among a plurality of users; A generating module, used for generating an iris recognition result according to the target feature data and the retrieved iris feature data; The extraction and classification module is specifically used for: extracting a plurality of stereoscopic iris image data from the standardized three-dimensional iris image data by using a three-dimensional image processing core of a certain size in the three-dimensional image processing layer; for each of the three-dimensional iris image data among the plurality of stereoscopic iris image data, performing vertical sectioning along the length, width and height directions of each of the three-dimensional iris image data, extracting three section data of corresponding sizes, and combining eight vertex data of each of the three-dimensional iris image data that have not been extracted with the center point data into oblique section data of corresponding sizes, so as to obtain a plurality of two-dimensional iris image data corresponding to the standardized three-dimensional iris image data.

8. The device according to claim 7, characterized in that The size of the 3D image processing core in the 3D image processing layer is 3×3×3; Correspondingly, the extraction and classification module is specifically used to: use the three-dimensional image processing layer to extract multiple three-dimensional iris image data with a size of 3×3×3 from the standardized three-dimensional iris image data; for each three-dimensional iris image data in the multiple three-dimensional iris image data, perform vertical sectioning along the length, width and height directions of each three-dimensional iris image data, extract three section data with a size of 3×3, and combine eight vertex data and center point data of each three-dimensional iris image data that have not been extracted into oblique section data with a size of 3×3, so as to obtain multiple two-dimensional iris image data corresponding to the standardized three-dimensional iris image data.

9. A computer device, characterized in that: The method comprises a processor and a memory for storing processor-executable instructions, wherein the processor implements the steps of the method according to any one of claims 1 to 6 when executing the instructions.

10. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the instructions are executed, the steps of the method according to any one of claims 1 to 6 are implemented.

Images