Three-dimensional iris reconstruction and expansion method

Abstract

The invention discloses a three-dimensional iris reconstruction and expansion method, and relates to the field of intelligent medical treatment, and the method comprises the following steps: 1, generating a three-dimensional iris label; step 2, an iris depth map prediction model; step 3, unfolding the iris; according to the method, iris depth information is accurately predicted by using an anatomical restriction relationship of a plurality of iris images in a video stream, three-dimensional iris reconstruction is realized, and the iris is expanded into a two-dimensional rectangular image and a depth map; according to the method, the defect that the depth of a traditional single image is insufficient is overcome, the iris image depth information is predicted by using multiple images of the video, and the iris texture between the iris images has an anatomical restriction relationship, so that the iris image depth information can be effectively predicted, the robustness of iris recognition is improved, and the method is suitable for various application scenes.

Description

Technical Field

[0001] The present invention relates to intelligent medical treatment, and particularly to a three-dimensional iris reconstruction and unfolding method. Background Art

[0002] Iris recognition, as a high-precision biometric technology, has been widely applied in fields such as identity authentication and security monitoring. Traditional iris recognition methods mainly rely on single two-dimensional iris images, which can only obtain the planar information of the iris and lack important depth information.

[0003] When iris images are acquired from different angles, since the iris itself is a three-dimensional structure, the two-dimensional images will present different morphologies. Although the rubber sheet model of Daugman is usually adopted for iris normalization, this model ignores the three-dimensional structure of the iris, which results in a significant decline in the performance of two-dimensional iris recognition in complex scenarios such as illumination changes, occlusion, and pose changes. Especially in the images taken from the side angle, the three-dimensional effect of the iris is very obvious.

[0004] In recent years, three-dimensional iris reconstruction methods based on optical coherence tomography (OCT) technology have been proposed. By irradiating the eye with specific structured light to obtain the three-dimensional model of the iris, and generating a planar graph and a depth graph for training a depth prediction model. However, the applicability of this method is limited. It is only applicable to iris images under specific structured light irradiation conditions, and does not fully utilize the anatomical constraint relationship of the iris texture in the video stream, which limits its universality and practical value.

[0005] With the development of deep learning technology, iris segmentation methods based on convolutional neural networks have emerged. These methods train network models through a large amount of labeled data and directly output segmentation results. However, these methods have the following limitations: 1. Simplify the iris as a two-dimensional structure, without considering the actual thickness and height of the iris, resulting in an overly simplified pupil modeling and a lack of anatomical constraints in pupil modeling; 2. The performance deteriorates under non-ideal conditions, is sensitive to view angle changes, the occlusion processing is not ideal, and the boundary positioning is inaccurate; 3. Manual labeling is time-consuming and laborious, the labeling results are inconsistent, lack anatomical verification, and it is difficult to guarantee the data quality; 4. The recognition accuracy is limited, the environmental adaptability is insufficient, the medical application value is limited, and the system robustness is insufficient.

[0006] Traditional mathematical methods rely on multi-view images (at least 2 images), based on geometric principles and triangulation, with a clear calculation process and strong interpretability. However, the accuracy depends on the image quality and camera calibration, and is sensitive to noise. Deep learning methods can estimate the depth from a single image, learn features based on data-driven, and implicitly learn depth cues, but the accuracy depends on the quality of the training data.

[0007] Advantages of traditional mathematical methods: solid theoretical foundation, no need for training data, and controllable and interpretable results; Disadvantages: multiple perspectives are required, calculations are complex, and sensitive to noise. Advantages of deep learning methods: single-image estimation is possible, fast speed, and strong robustness; Disadvantages: a large amount of training data is required, limited generalization, and difficult to interpret results. If depth information is obtained from a single image, there is a certain dependence on light and shadow. Using multiple images to obtain depth information will yield better results, but the technical difficulty will increase, and key point extraction and calibration are required.

[0008] Recent research has shown that incorporating three-dimensional information can significantly improve the performance of biometric systems. Specifically in iris recognition, existing research has attempted to reconstruct three-dimensional iris models from multiple two-dimensional images or use specialized devices to estimate depth information. However, these methods often require complex imaging setups or multiple image captures, limiting their practical applications.

[0009] The main challenges faced by iris-based biometric systems include: 1. Limited information obtained from traditional two-dimensional iris images, unable to consider the natural curvature and depth changes of the iris. 2. Performance degradation under different lighting conditions, viewing angles, and pupil dilation states. 3. Difficulty in accurately segmenting the iris region and normalizing it for feature extraction. 4. Lack of effective methods for reconstructing the three-dimensional structure of the iris from limited two-dimensional imaging data.

[0010] In the field of three-dimensional eyeball modeling, existing research mainly focuses on eyeball movement modeling and gaze estimation, with less consideration of the three-dimensional structural characteristics of the iris. Traditional eyeball models usually adopt a single spherical structure, which fails to accurately reflect the thickness and height characteristics of the iris, resulting in the loss of important three-dimensional information during iris segmentation and feature extraction.

[0011] The three-dimensional iris reconstruction and unfolding technology has important application values in the fields of medical diagnosis, biometrics, or intelligent medical systems. In the field of medical diagnosis, by reconstructing the fine structure of the iris, it can assist in diagnosing diseases such as iris defects, iris cysts, and glaucoma (abnormal iris morphology); or provide high-precision three-dimensional models for iris laser surgery and artificial iris implantation to improve surgical safety; or quantify parameters such as iris thickness and blood vessel distribution to study the progression of diseases (such as diabetic iris disease). In the field of biometrics, iris texture is unique and stable. The standardized image after unfolding can extract stable features (such as Gabor filter coding) for high-precision identity recognition; by reconstructing the three-dimensional features of the iris (such as microvascular structure), it can distinguish real irises from forged images (such as contact lens imitations). In the field of intelligent medical system integration, in medical scenarios, iris recognition can replace traditional IDs to ensure the accurate association of patients with electronic medical records; the feature coding after iris unfolding is irreversible, avoiding the leakage of original biological information and meeting the requirements of medical data security. However, most of the existing technologies currently use single iris images, which only include two-dimensional information and lack depth information. Some research is based on OCT to obtain a three-dimensional iris model, generate an iris plan view and a depth map according to the three-dimensional iris model, and train a depth model based on this to predict the depth map information of iris images. Its principle is to use specific structured light to irradiate the eye to obtain iris photos, but its defect is that it is only applicable to iris images irradiated by structured light, lacks universality, and does not utilize the anatomical constraint relationship of video stream information. The market needs a three-dimensional iris reconstruction and unfolding method that can accurately utilize multiple images of a video to predict the depth information of iris images and improve the recognition accuracy and robustness. Summary of the Invention

[0012] To solve the problems in the existing iris recognition technology, such as insufficient information in two-dimensional methods and difficulty in dealing with complex scenarios, as well as the limited applicability of existing three-dimensional reconstruction methods and the failure to fully utilize video information, the purpose of the present invention is to propose a more accurate and robust three-dimensional iris reconstruction and unfolding method, aiming to better understand and process the three-dimensional structure of the iris, significantly improve the accuracy and robustness of iris recognition, and thus enhance the application effects in the fields of three-dimensional eye movement detection, iris recognition, security authentication, medical diagnosis, etc.

[0013] To achieve the above objectives, the present invention adopts the following innovative technical solutions: A three-dimensional iris reconstruction and unfolding method, including the following: Step 1, generation of three-dimensional iris labels: 1) Three-dimensional iris model based on OCT: Using OCT technology to obtain a high-precision three-dimensional iris model, and simulating to generate an iris image video stream and its corresponding depth map labels; 2) 3D iris model based on eye movement video: Based on the iris segmentation results of the eye movement video, construct an eyeball model, obtain parameters such as the eyeball center, iris radius, and eyeball radius, and calculate the optical axis and the parametric equation of the iris circle; Step 2, Iris depth map prediction model: 1) Construct a deep learning model with the input being a sequence of temporal iris images and the output being the corresponding iris depth map. By utilizing the information correlation between consecutive frames in the video stream, the deficiencies of single-image depth estimation are overcome.

[0014] 2) Use a CNN encoder to extract multi-level features of each frame of iris image, utilize an LSTM module to fuse the temporal information in the video sequence, and introduce a multi-scale attention mechanism to enable the model to focus on image regions and features that are more important for depth prediction.

[0015] 3) Combine an anatomical constraint loss function: During the model training process, introduce constraints based on the anatomical structure of the eyeball as part of the loss function, such as prior knowledge of the eyeball center, iris radius, eyeball radius, etc. This helps to improve the accuracy of depth prediction and the rationality of conforming to the real iris structure, and overcomes the prediction results that may not conform to the anatomical structure generated by traditional deep learning methods. Step 3, Iris unfolding: 1) Construct a 3D iris image based on the 2D image and depth information of the iris; 2) Unfold the 3D iris into a 2D rectangular image and a depth map. The specific steps include constructing a 3D coordinate system, calculating the rotation matrix, adjusting the iris to the frontal position, and performing polar coordinate transformation.

[0016] For the aforementioned 3D iris reconstruction and unfolding method, the specific content of Step 1, 3D iris label generation includes: 1) 3D iris model based on OCT: Utilize OCT technology to obtain a high-precision 3D iris model, and simulate and generate an iris image video stream and its corresponding depth map label; 2) 3D iris model based on eye movement video: Parameter acquisition: Based on the iris segmentation results of the eye movement video, construct an eyeball model and obtain the eyeball center , iris radius r, eyeball radius R, and iris center ; Optical axis calculation: Optical axis , unit optical axis ; Parametric equation of the iris circle: ; Feature extraction and matching: Use the SIFT / SURF algorithm to extract iris texture features, calculate the three-dimensional coordinates of feature points, and generate depth map labels; the iris texture features include: pupil, pupillary ring, contraction furrow, radial striations, crypts, iris freckles, and Wolfflin nodules.

[0017] The aforementioned three-dimensional iris reconstruction and unfolding method, step two, the specific method of the iris depth map prediction model includes: 1) Construct a deep learning model, with the input being a sequence of temporal iris images and the output being the corresponding iris depth map; 2) Use a CNN encoder to extract features of temporal iris images, an LSTM module to fuse temporal information, and a depth decoder to generate an iris depth map, combined with an anatomical constraint loss function and a multi-scale attention mechanism; The aforementioned anatomical constraint loss function is: This loss function constrains the predicted iris radius, eye center, eye radius, and the distance from iris feature points to the eye center, thereby using the geometric structure information of the eye to standardize the prediction of the iris depth map; among them, the eye center , iris radius and eye radius are estimated based on the iris segmentation results of eye movement videos, where is the weight used to balance the importance of the anatomical constraint term; Assume that N iris feature points are extracted, the predicted three-dimensional coordinates of the i-th feature point are , the predicted eye center is , and there is an expected distance ; then, ; The multi-scale attention mechanism is: Extract multi-scale features by using convolutional kernels of different sizes, and introduce an attention mechanism during the decoding process to focus on key iris texture regions. First, pre-train on OCT simulation data, and then fine-tune on real eye movement video data, combined with data augmentation to simulate scenarios of illumination, occlusion, and pupil scaling.

[0018] For the aforementioned three-dimensional iris reconstruction and unfolding method, the sequence of temporal iris images is at least 3 frames.

[0019] For the aforementioned three-dimensional iris reconstruction and unfolding method, the iris unfolding method specifically includes: Three-dimensional iris reconstruction: Construct a three-dimensional iris image through the two-dimensional image of x and y axis information and the depth information of the z axis; Unfolding process: 1) Construct a three-dimensional coordinate system with the eye center as the origin.

[0020] 2) Calculate the rotation matrix based on the iris center and the eyeball center vector , and adjust the iris to the frontal position.

[0021] 3) Apply the rotation matrix to each feature point of the iris texture , and adjust it to the frontal position.

[0022] 4) Through polar coordinate transformation, expand the iris circle into a rectangular texture image and a depth map.

[0023] Parametric expansion: 1) Texture expansion: Map the iris annular texture into a rectangular image.

[0024] 2) Depth preservation: Synchronously expand the depth information to generate a corresponding depth map.

[0025] For the foregoing three-dimensional iris reconstruction and expansion method, the information acquisition module for collecting OCT images or eye movement videos is a head-mounted device.

[0026] For the foregoing three-dimensional iris reconstruction and expansion method, the method is applied to eye movement tracking, biometric authentication, medical diagnosis, or computer vision and image processing. Further, it is particularly applicable to applications with high-precision and high-robustness requirements in the following fields: three-dimensional eyeball movement detection, high-security biometric authentication, refined ophthalmic medical diagnosis and surgical planning.

[0027] The beneficial effects of the present invention are as follows: The present invention proposes to use multiple pictures of a video to predict the depth information of an iris image. Since the camera of the head-mounted device and the head are relatively fixed, and the eyeball center remains stationary during eyeball movement, there is an anatomical constraint relationship between the iris textures of each iris picture. Based on this, the depth information of the iris image can be effectively predicted, overcoming the limitation of insufficient depth of a single image.

[0028] The present invention introduces anatomical constraints into the deep learning model to improve the accuracy and robustness of depth prediction.

[0029] The present invention proposes a method for expanding a three-dimensional iris into a two-dimensional rectangular image and a depth map, which is convenient for subsequent iris recognition and analysis. Description of the Drawings

[0030] Figure 1 is a schematic diagram of a single iris plan view and a depth map in the prior art (the iris plan view is above and the iris depth map is below); Figure 2 is the eyeball model and the optical axis iris depth map of the present invention; Figure 3Schematic diagram of iris texture features of the present invention; (Pupil refers to the pupil, Collarette refers to the pupillary ring, Contraction furrows refers to the contraction grooves, Radial streaks refers to the radial streaks, Crypts refers to the crypts, Iris Freckles refers to the iris freckles, Wolfflin nodules refers to the Wolfflin nodules); Figure 4 Iris segmentation result based on the constraints of the three-dimensional eyeball model parameters of the present invention (Initial irisLabels: Initial iris labels, 3D Eyeball Model: 3D eyeball model, Iris Annotation: Iris annotation, Iris center: Iris center, Powell optimiza: Powell optimization, Eye ball center: Eyeball center, IrisRadius: Iris radius, Initial iris: Initial iris, Iris ellip: Iris ellipse); Figure 5 Schematic diagram of iris texture feature extraction and matching of the present invention (frontal position); Figure 6 Schematic diagram of iris texture feature extraction and matching of the present invention (optical axis movement detection); Figure 7 Three-dimensional iris unfolding diagram of the present invention (Pupil / Iris Boundary refers to the pupil / iris boundary, Iris / Sclera Boundary refers to the iris / sclera boundary); Figure 8 Stereogram of the iris circle of the present invention unfolded into a rectangle. Specific implementation manner

[0031] The present invention will be specifically introduced below in conjunction with the accompanying drawings and specific embodiments.

[0032] A three-dimensional iris reconstruction and unfolding method includes the following contents: Step 1, generation of three-dimensional iris labels: 1.1 Three-dimensional iris model based on OCT: Using OCT technology to obtain a high-precision three-dimensional iris model, simulating and generating an iris image video stream and its corresponding depth map labels; Based on OCT, obtain a three-dimensional iris model, as Figure 2 shown, simulate and construct an iris image video stream, as well as an iris depth map label. The original method only established a single iris picture and its corresponding iris depth map, as Figure 1 shown, the upper part is the iris plane diagram, and the lower part is the iris depth map. The present invention provides high-precision three-dimensional ground truth for supervising the training of the depth prediction model.

[0033] 1.2 3D Iris Model Based on Eye Movement Videos: Based on the iris segmentation results of eye movement videos, construct an eyeball model, obtain parameters such as the eyeball center, iris radius, and eyeball radius, and calculate the optical axis and the parametric equation of the iris circle: 1.2.1 Parameter Acquisition: Based on the iris segmentation results of eye movement videos, construct an eyeball model and obtain the eyeball center , iris radius r, eyeball radius R, and iris center ; Based on the iris segmentation labels of eye movement videos, construct an eyeball model, including the eyeball center, iris radius, and eyeball radius The eyeball center remains stationary during eyeball movement. Let the eyeball center be [x, y, 0], the iris radius be r, the eyeball radius be R, and the iris center Iris be (x0, y0, z0). Then, the optical axis can be determined based on the eyeball center and the iris center. Optical Axis Calculation: The optical axis , unit optical axis ; Parametric Equation of the Iris Circle: ; 1.2.2 Eyeball Movement Modeling: Modeling is performed based on the following parameters: eyeball center C = (x, y, 0), origin of the coordinate system; Iris center I = (x0, y0, z0) Optical axis vector Parametric Equation of the Iris Circle: ; The iris segmentation results constrained by the 3D eyeball model parameters are as Figure 4 shown.

[0034] 1.2.3 Feature Extraction and Matching: Use the SIFT / SURF algorithm to extract iris texture features, calculate the 3D coordinates of feature points, and generate depth map labels; Iris texture features include: pupil, pupil ring, contraction groove, radial stripes, crypts, iris freckles, and Wolfflin nodules. The iris texture features are as Figure 3 shown.

[0035] 1.2.3.1 The frontal position is as Figure 5 shown. The schematic diagram of iris texture extraction and matching and optical axis movement detection is as Figure 6 shown.

[0036] 1.2.3.2 Calculation of the 3D Coordinates of Iris Feature Points Set of iris feature points The i-th feature point Pi = Calculate its depth information based on the rotation matrix between the feature points of each figure in the eye movement video and the frontal position.

[0037] 1.2.3.3 Establish iris depth map labels ```python # Pseudocode example: depth map generation for frame in video_frames: estimate_pose(eye_center, iris_radius) compute_rotation_matrix(alpha, beta) project_3d_points_to_depth_map(feature_points) Step 2, Iris depth map prediction model: 2.2.1 Build a deep learning model, with the input being a sequence of temporal iris images (at least 3 frames), and the output being the corresponding iris depth map; 2.2.2 Use a CNN encoder to extract image features, an LSTM module to fuse temporal information, and combine an anatomical constraint loss function and a multi-scale attention mechanism; The following demonstrates the specific process where the input is three iris planar graphs and the output is three iris depth maps.

[0038] 2.2.2.1 Network architecture Input: Sequence of temporal iris images (more than 3 frames) Backbone network: ```mermaid graph TD Input1 --> CNN_Encoder Input2 --> CNN_Encoder Input3 --> CNN_Encoder CNN_Encoder --> LSTM_Fusion LSTM_Fusion --> Depth_Decoder ``` 2.2.2.2 The anatomical constraint loss function is: ; 2.2.2.3 The multi-scale attention mechanism enhances texture feature extraction: Two-stage training: First pre-train on OCT simulation data, and then fine-tune on real video data; Data augmentation: Simulate different lighting conditions, occlusions, and pupil dilation states. 3. Step 3, Iris unfolding: 3.1 From the 2D iris image (x-axis and y-axis information) and depth information (z-axis information), a 3D iris image can be constructed to obtain 3D iris texture information; 3.2 Unfold the 3D iris into a 2D rectangular image and a depth map. The specific steps include constructing a 3D coordinate system, calculating the rotation matrix, adjusting the iris to the frontal position, and performing polar coordinate transformation.

[0039] The unfolding of the iris is to unfold the three-dimensional solid iris in 3D space, including a 2D rectangular image (x-axis and y-axis information) and depth information (z-axis information), and a 3D cuboid relief structure can be constructed. The specific unfolding process: 3.2.1 Construct a 3D space coordinate system with the center of the eyeball as the origin 3.2.2 Calculate the rotation matrix p for rotating to the frontal position based on the vector between the iris center and the eyeball center.

[0040] 3.2.3 Apply the rotation matrix p to the 3D iris circle in 3D space to rotate it to the frontal position 3.2.4 Each feature point vector of the iris texture includes (x, y, z) information. Apply the rotation matrix p to each feature point vector to rotate it to the frontal position 3.2.5 Unfold the iris circle, including a 2D rectangular image (x-axis and y-axis information) and depth information (z-axis information), as well as a 3D cuboid relief. The 3D iris unfolded image is as Figure 7 and Figure 8 shown.

[0041] Definition of parametric unfolding: The iris is an annular region, usually located on a 3D surface (such as the surface of the eyeball). The purpose of parametric unfolding is: Texture unfolding: Map the annular texture of the iris (the image after 2D projection) from a circular ring shape to a rectangular image. Depth preservation: During the unfolding process, preserve the depth information (z-axis value) of each pixel point in 3D space to generate a depth map corresponding to the unfolded rectangular texture image.

[0042] 4. Usage example: 4.1. Input data: 3D iris data (e.g., in point cloud format, containing x, y, z coordinates).

[0043] Iris center coordinates, inner and outer radii, target dimensions (width and height) of the unfolded image.

[0044] 4.2. Projection: Project the 3D iris onto a 2D plane to generate a 2D texture image and a corresponding z-value map.

[0045] 4.3. Unwrapping: Perform polar coordinate transformation on the two - dimensional texture image to generate a rectangular texture image.

[0046] Perform the same polar coordinate transformation on the z - value map to generate a rectangular depth map.

[0047] 4.4. Output: The unwrapped rectangular texture image.

[0048] The corresponding rectangular depth map.

[0049] 4.5 Example code: def unwrap_iris_with_depth(img3D, center, r_min, r_max, width,height): img2D = project_3d_to_2d(img3D) # Project to two - dimensions z_values = extract_z_values(img3D) # Extract depth values texture = cv2.warpPolar(img2D, (width, height), center, r_max,cv2.WARP_INVERSE_MAP + cv2.INTER_CUBIC) z_map = cv2.warpPolar(z_values, (width, height), center, r_max,cv2.WARP_INVERSE_MAP + cv2.INTER_CUBIC) return texture, z_map.

[0050] 5. Application of this method 5.1 Enhancing biometric authentication The proposed 3D iris reconstruction method greatly improves the efficiency of biometric authentication in the following ways: 1. Improve the recognition accuracy, especially under challenging conditions such as different lighting and perspectives 2. Enhance the resistance to spoofing attempts that rely on two - dimensional iris representations 3. Facilitate more accurate iris feature matching by combining depth information.

[0051] 5.2 Medical diagnosis In the medical field, our method provides valuable functions for the following aspects: 1. More accurately evaluate the condition of the iris and anterior segment of the eye.

[0052] 2. Track subtle changes in the iris structure, which may indicate the progression of diseases 3. Support surgical planning and evaluation in ophthalmic surgery 4. Computer vision and image processing.

[0053] 5.3 Computer vision and image processing: The three-dimensional iris reconstruction technology can be widely applied to applications with high-precision and high-robustness requirements in the following fields: The method is applied to eye movement tracking, biometric authentication, medical diagnosis or computer vision and image processing, and is particularly suitable for applications with high-precision and high-robustness requirements in the following fields: three-dimensional eye movement detection, high-security biometric authentication, refined ophthalmic medical diagnosis and surgical planning.

[0054] 1. Three-dimensional eye movement detection: By reconstructing the real three-dimensional structure of the iris, the deformation influence caused by traditional two-dimensional methods when the viewing angle changes is overcome, and the movement trajectory of the iris in three-dimensional space can be tracked more accurately.

[0055] Combined with the depth prediction of the sequential iris image sequence, the depth change of eye movement can be analyzed to provide more comprehensive eye movement information.

[0056] The unfolded iris image and depth map provide a standardized data format for subsequent eye movement analysis and modeling.

[0057] 2. High-security biometric authentication: Using the reconstructed three-dimensional iris structure and the unfolded depth information, more abundant and robust iris features can be extracted, improving the accuracy and anti-spoofing ability of identity recognition. For example, the iris micro-structure and depth information that are difficult to capture in two-dimensional images can be analyzed.

[0058] The unfolded feature encoding is irreversible, ensuring the security of the original biological information, and is particularly suitable for scenarios with strict requirements for data security.

[0059] 3. Refined ophthalmic medical diagnosis and surgical planning: Through high-precision three-dimensional iris reconstruction, the morphological features of the iris can be evaluated more accurately, assisting in the diagnosis of eye diseases related to the three-dimensional structure of the iris, such as iris defects, iris cysts, and glaucoma.

[0060] Provide an accurate three-dimensional model for surgeries such as iris laser surgery and artificial iris implantation, helping doctors to perform more refined surgical planning and improving the success rate and safety of surgeries.

[0061] Quantitatively analyzing three-dimensional parameters such as the thickness and blood vessel distribution of the iris can provide an objective basis for studying the progression and efficacy evaluation of eye diseases. The present invention proposes to use multiple pictures of a video to predict the depth information of an iris image. Since the camera of the head-mounted device and the head are relatively fixed, and the center of the eyeball remains stationary during eye movement, there is an anatomical constraint relationship between the iris textures in each iris picture. Based on this, the depth information of the iris image can be effectively predicted. The method of the present invention overcomes the defect of insufficient depth of traditional single images, improves the robustness of iris recognition, is applicable to various application scenarios, and has significant innovation and practical value.

[0062] The foregoing has shown and described the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the above embodiments do not limit the present invention in any form. Any technical solutions obtained by means of equivalent replacement or equivalent transformation fall within the protection scope of the present invention.

Claims

1. A three-dimensional iris reconstruction and unfolding method, characterized in that: It includes the following: Step 1: 3D iris label generation: 1) 3D iris model based on frequency domain optical coherence tomography (OCT): Use OCT technology to obtain a high-precision 3D iris model, and simulate and generate iris image video streams and their corresponding depth map labels; 2) 3D iris model based on eye movement video: Based on the iris segmentation results of the eye movement video, the eyeball model is constructed to obtain parameters such as the eyeball center, iris radius, corneal curvature and eyeball radius, and the optical axis and iris circle parameter equations are calculated; Step 2: Iris depth map prediction model: 1) Build a deep learning model with the input as a time-series iris image sequence and the output as the corresponding time-series iris depth map; 2) A CNN encoder is used to extract image features, and an LSTM module is used to fuse temporal information, combined with an anatomical constraint loss function that at least includes a constraint on the difference between the predicted iris radius and the eyeball center, and a multi-scale attention mechanism; Step 3: Iris expansion: 1) Using the 2D image and depth information of the iris, a corresponding depth value is assigned to each pixel of the 2D image to construct a 3D iris image; 2) Expand the three-dimensional iris into a two-dimensional rectangular image and depth map. The specific steps include constructing a three-dimensional coordinate system, calculating a rotation matrix, adjusting the iris to the frontal position, and performing a polar coordinate transformation based on relaxing geometric constraints.

2. A three-dimensional iris reconstruction and unfolding method according to claim 1, characterized in that: The specific contents of step 1, generating a three-dimensional iris tag, include: 1) OCT-based 3D iris model: Use OCT technology to obtain a high-precision 3D iris model, and simulate and generate iris image video streams and their corresponding depth map labels under different pupil sizes and lighting conditions; 2) 3D iris model based on eye movement video: Parameter acquisition: Based on the iris segmentation results of the eye movement video, the eyeball model is constructed to obtain the eyeball center including the corneal curvature , iris radius r, eyeball radius R and iris center ; Optical axis calculation: Optical axis , unit optical axis ; Iris circle parameter equation: ； Feature extraction and matching: Use SIFT / SURF and other algorithms to extract iris texture features, calculate the three-dimensional coordinates of feature points through feature matching, and generate depth map labels; the iris texture features include: pupil, pupil ring, contraction groove, radial stripes, crypts, iris freckles and Wolfflin's nodules.

3. The three-dimensional iris reconstruction and unfolding method according to claim 1, characterized in that: In step 2, the specific method of the iris depth map prediction model includes: 1) Build a deep learning model with the input as a time-series iris image sequence and the output as the corresponding time-series iris depth map; 2) A CNN encoder containing multiple convolutional layers and pooling layers is used to extract the time series iris image features. The LSTM module is used to model the motion and deformation information of the iris in the time series. A deep decoder containing deconvolution layers or upsampling layers is used to generate the iris depth map, combined with the anatomical constraint loss function and multi-scale attention mechanism; The anatomical constraint loss function is: The loss function constrains the predicted iris radius, eyeball center, eyeball radius, and the distance from the iris feature point to the eyeball center, thereby using the geometric structure information of the eyeball to standardize the prediction of the iris depth map; among them, the eyeball center , iris radius and eyeball radius The iris segmentation result based on the eye movement video is estimated, where: is the weight used to balance the importance of anatomical constraints; Assuming that N iris feature points are extracted, the predicted 3D coordinates of the i-th feature point are , the predicted eye center is , and there is a desired distance ;So, ; The multi-scale attention mechanism is as follows: multi-scale features are extracted by using convolution kernels of different sizes, and an attention mechanism is introduced in the decoding process to focus on the key iris texture area. It is first pre-trained on OCT simulation data, and then fine-tuned on real eye movement video data, combined with data enhancement to simulate scenes such as lighting, occlusion, and pupil zoom.

4. A three-dimensional iris reconstruction and unfolding method according to claim 3, characterized in that: The temporal iris image sequence is at least 3 frames, so as to use the anatomical constraint relationship of iris texture between consecutive frames to perform more accurate depth information prediction.

5. The three-dimensional iris reconstruction and unfolding method according to claim 1, characterized in that: The iris expansion method specifically comprises: 3D iris reconstruction: construct a 3D iris image through the x-axis and y-axis information of the 2D iris image and the z-axis information of the corresponding depth information map; Unfolding process: 1) Construct a three-dimensional coordinate system with the center of the eyeball as the origin; 2) Calculate the rotation matrix based on the iris center and eyeball center vectors , adjust the iris to the emmetropia position; 3) Apply the rotation matrix to each feature point vector (containing x, y, z information) of the 3D iris texture , so that it is converted to the normal view position; 4) Through polar coordinate transformation, the three-dimensional iris circle at the frontal position is expanded into a two-dimensional rectangular image containing texture information (x-axis and y-axis information) and the corresponding depth map (z-axis information); Parametric expansion: 1) Texture unfolding: Mapping the annular texture of the iris from the two-dimensional projected image to a rectangular image; 2) Depth preservation: While the texture is being expanded, the depth information (z-axis value) of each pixel in the three-dimensional space is retained to generate a depth map corresponding to the expanded rectangular texture image; the expanded two-dimensional rectangular image and depth map can be used to construct a three-dimensional rectangular relief structure.

6. The three-dimensional iris reconstruction and unfolding method according to claim 1, characterized in that: The information acquisition module for acquiring OCT images or eye movement videos is a head-mounted device.

7. The three-dimensional iris reconstruction and unfolding method according to claim 1, characterized in that: The method is applied to eye tracking, biometric identity authentication, medical diagnosis or computer vision and image processing, preferably to applications with high precision and high robustness requirements in the following fields: three-dimensional eye movement detection, high-security biometric identity authentication, refined ophthalmic medical diagnosis and surgical planning.

Images