A personalized eyeglass frame design method and system based on 3D facial features

Abstract

The application provides a personalized glasses frame design method and system based on 3D facial features. The method obtains facial 3D geometric data through a scanning device; extracts key feature points using an improved HRNet architecture; inputs the feature points into a multi-layer perception machine model to generate an initial frame parameter set containing 32 core parameters of four parts, i.e., frame geometry, nose bridge, temple, and overall design; constructs a multi-dimensional constraint matrix composed of five sub-matrices, i.e., geometry, style, manufacturing, comfort, and medical; establishes a constraint optimization objective function, taking the initial parameter set as the decision variable and the multi-dimensional constraint matrix as the constraint condition, and iteratively solves the optimal frame parameter set with the least constraint condition violation degree through a sequential quadratic programming algorithm; and finally generates a 3D model of the personalized frame. The application realizes full-process automatic design, ensures high adaptation of the frame to the patient's face, and is especially suitable for medical rehabilitation scenarios such as facial asymmetry or postoperative repair.

Description

Technical Field

[0001] This application relates to the field of personalized eyeglass frame design, and more specifically, to a method and system for designing personalized eyeglass frames based on 3D facial features. Background Technology

[0002] With the increasing application of computer-aided design (CAD) and additive manufacturing technologies in the field of precision medicine, the eyewear industry is undergoing a paradigm shift from "standardized mass production" to "high-precision customization." For groups requiring special vision correction or facial reconstruction, eyeglass frames are not only carriers of optical lenses, but also important components for facial rehabilitation and anatomical structural support.

[0003] Currently, eyeglass frame designs on the market still largely employ standardized size systems and limited options for minor style adjustments. In the actual design and fitting process, professionals primarily rely on manual measurements (such as pupillometers and calipers) and their personal intuition for judgment. Although some high-end custom services have begun to incorporate 3D scanning technology, most of these efforts remain at the level of visual "virtual try-on," lacking a digital foundation for understanding the deep mapping relationship between facial anatomical features and frame geometric parameters.

[0004] Therefore, this application provides a personalized eyeglass frame design method and system based on 3D facial features to solve one of the above-mentioned technical problems. Summary of the Invention

[0005] The purpose of this application is to provide a method and system for designing personalized eyeglass frames based on 3D facial features, which can solve at least one of the aforementioned technical problems. The specific solution is as follows: According to a specific embodiment of this application, in a first aspect, this application provides a method for designing personalized eyeglass frames based on 3D facial features, comprising: Facial scan data of the user is acquired using a scanning device; wherein the facial scan data includes at least facial 3D geometric data; key feature points are extracted from the facial 3D geometric data using an improved HRNet architecture; the key feature points are used as input and mapped through a multilayer perceptron model to generate an initial frame parameter set containing 32 core parameters; wherein the 32 core parameters consist of four parts including frame geometric parameters, bridge of the nose parameters, temple parameters, and overall design parameters; a multi-dimensional constraint matrix is ​​constructed, consisting of geometric constraint sub-matrices, style constraint sub-matrices, manufacturing constraint sub-matrices, comfort constraint sub-matrices, and medical constraint sub-matrices; a constraint optimization objective function is established, using the initial frame parameter set as decision variables and the multi-dimensional constraint matrix as constraints, and an iterative solution using a sequential quadratic programming algorithm is used to find the optimal frame parameter set that minimizes the violation of each dimension constraint condition; a personalized frame 3D model is generated based on the optimal frame parameter set.

[0006] In one embodiment, the key feature points are 22 feature points located at different positions on the face, including: the left mandibular endpoint, right mandibular endpoint, left masseter point, right masseter point, left temporal point, right temporal point, mandibular projection point, and tangent point distributed in the facial contour area; the left outer corner of the eye, right outer corner of the eye, left inner corner of the eye, right inner corner of the eye, left brow peak point, and right brow peak point distributed in the eye area; the tip of the nose, root of the nose, left ala of the nose, and right ala of the nose distributed in the nose area; and the left corner of the mouth, right corner of the mouth, center point of the upper lip, and center point of the lower lip distributed in the mouth area.

[0007] In one embodiment, the core parameters include at least the frame width, frame height, bridge width, nasal bridge width, and temple length, which are mapped as follows: the frame width is mapped based on the Euclidean distance between the left and right mandibular endpoints; the frame height is mapped based on the distance between the nasal tip and the left brow peak; the bridge width is mapped based on the distance between the left and right alar points combined with a preset tissue repair factor; the nasal bridge width is mapped based on the distance between the left and right alar points combined with the width change; and the temple length is mapped based on the distance between the left temporal point and the mandibular projection point combined with a length compensation value.

[0008] In one embodiment, the method further includes: calculating the reliability score of each of the key feature points; wherein the reliability score is obtained by weighted summation of the confidence score, consistency score and spatial coherence score extracted from the key feature points; and adjusting the original weights of each parameter in the optimal frame parameter set according to the reliability score.

[0009] In one implementation, during the iterative solution process, a three-level priority mechanism is used to handle constraint conflicts. The priority of different constraints is defined as follows: geometric feasibility, structural strength, and manufacturing feasibility are set as the first priority hard constraints that cannot be violated; aesthetic requirements, style consistency, and comfort are set as the second priority soft constraints that can be moderately violated; and weight minimization, cost minimization, and production efficiency are set as the third priority target constraints. When a constraint conflict occurs, parameters are adjusted first to satisfy the first priority hard constraints, and the second priority soft constraints are handled by minimizing the degree of violation, provided that the first priority hard constraints are satisfied.

[0010] In one embodiment, the method further includes: constructing a finite element model of the eyeglass frame based on the optimal frame parameter set, and reconstructing a facial finite element model containing the biomechanical properties of skin, subcutaneous fat, and cartilage tissue using the facial scan data; applying a self-weight load, facial muscle force based on electromyography data, and periodic load based on blink frequency to the finite element model of the eyeglass frame; determining the von Mises stress and contact pressure distribution in the contact area between the finite element model of the eyeglass frame and the facial finite element model; identifying the intra-domain pressure in the nasal bridge region, temporal bone region, and mandibular border region based on the von Mises stress and the contact pressure distribution, and determining whether the intra-domain pressure in each region exceeds a preset pain threshold; if there is a target intra-domain pressure exceeding the preset pain threshold, performing geometric optimization on the region corresponding to the target intra-domain pressure in the finite element model of the eyeglass frame; wherein, the geometric optimization includes increasing the corner radius and / or adjusting the wall thickness distribution.

[0011] In one embodiment, before determining the von Mises stress and the contact pressure distribution, the method further includes: calculating the natural frequency of the frame structure characterized by the finite element model of the frame; if the natural frequency is within the blink frequency range, adjusting the parameters of the finite element model of the frame until the natural frequency avoids the blink frequency range.

[0012] In one embodiment, generating a 3D model of a personalized eyeglass frame based on the optimal frame parameter set includes: converting the geometrically optimized finite element model of the frame into a 3D printing format, performing slicing optimization according to additive manufacturing processes, and generating a support structure.

[0013] In one embodiment, the method further includes: increasing the weight of the geometric constraint submatrix in response to detecting facial asymmetry; and increasing the weight of the comfort constraint submatrix and the medical constraint submatrix in response to detecting that the user is in the early postoperative recovery period.

[0014] According to a specific embodiment of this application, in a second aspect, this application provides a personalized eyeglass frame design system based on 3D facial features, comprising: The system includes a data acquisition module for acquiring facial scan data from a user using a scanning device; the facial scan data includes at least 3D facial geometric data. A feature extraction module extracts key feature points from the 3D facial geometric data using an improved HRNet architecture. A parameter generation module takes the key feature points as input and maps them through a multilayer perceptron model to generate an initial frame parameter set containing 32 core parameters; these 32 core parameters consist of four parts: frame geometric parameters, bridge of the nose parameters, temple parameters, and overall design parameters. A constraint optimization module constructs a multi-dimensional constraint matrix consisting of geometric constraint sub-matrices, style constraint sub-matrices, manufacturing constraint sub-matrices, comfort constraint sub-matrices, and medical constraint sub-matrices. A finite element optimization module establishes a constraint optimization objective function, using the initial frame parameter set as decision variables and the multi-dimensional constraint matrix as constraints, iteratively solving for the optimal frame parameter set that minimizes the violation of constraints in each dimension using a sequential quadratic programming algorithm. A manufacturing implementation module generates a 3D model of a personalized frame based on the optimal frame parameter set.

[0015] Compared with the prior art, the above-described solution of this application has at least the following beneficial effects: This application provides a personalized eyeglass frame design method based on 3D facial features. Objective facial data is obtained through 3D scanning, feature points are accurately extracted using HRNet, and then mapped by a multilayer perceptron to generate a 32-dimensional initial parameter set covering the frame, bridge, temples, and overall design, achieving a scientific transformation from facial morphology to design parameters. Based on this, a five-dimensional constraint matrix of geometry, style, manufacturing, comfort, and medical aspects is constructed, and a sequential quadratic programming algorithm is used for multi-objective optimization to ensure that the final parameters achieve optimal comprehensive performance while satisfying all hard constraints. This method achieves fully automated design throughout the entire process, significantly improving design efficiency and consistency, and ensuring a high degree of fit between the frame and the patient's face and wearing comfort. It is particularly effective in providing precise and feasible customized solutions for individuals with special medical needs such as facial asymmetry or post-operative repair. Attached Figure Description

[0016] Figure 1 A flowchart illustrating a personalized eyeglass frame design method based on 3D facial features is shown. Figure 2 A block diagram of a personalized eyeglass frame design system based on 3D facial features according to an embodiment of this application is shown. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the application. The singular forms “a,” “said,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0019] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0020] It should be understood that although the terms first, second, third, etc., may be used in the embodiments of this application, these descriptions should not be limited to these terms. These terms are only used to distinguish the descriptions. For example, first may also be referred to as second without departing from the scope of the embodiments of this application, and similarly, second may also be referred to as first.

[0021] Depending on the context, the words “if” or “suppose” as used here can be interpreted as “when” or “in response to determination” or “in response to detection.” Similarly, depending on the context, the phrases “if determination” or “if detection (of the stated condition or event)” can be interpreted as “when determination” or “in response to determination” or “when detection (of the stated condition or event)” or “in response to detection (of the stated condition or event).”

[0022] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a product or system comprising a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a product or system. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the product or system that includes that element.

[0023] It should be noted that any symbols and / or numbers present in the specification that are not marked in the accompanying drawings are not reference numerals.

[0024] In practical applications, especially in medical rehabilitation scenarios, the following technical problems urgently need to be addressed in related technologies: (1) Extremely low anatomical fit: Standardized frames cannot be individually adjusted to the micron level according to the patient's specific facial anatomy. Especially in cases of facial trauma, asymmetry, or postoperative tissue loss, standard frames are difficult to fix precisely, often resulting in tilting, slipping, or fitting failure.

[0025] (2) Lack of systematization and repeatability in the design process: Traditional design processes rely too much on human experience, making it difficult to accurately quantify the relationship between key facial features (such as nasal bridge slope and temporal curvature) and frame parameters. This not only leads to low design efficiency but also results in extremely poor consistency of design results, making it impossible to achieve batch-based precise iteration for rehabilitation cycles.

[0026] (3) Lack of biomechanical comfort: Existing technologies often only pursue "geometric fit" while ignoring the dynamic pressure interaction between the frame and facial soft tissues (such as skin, fat, and cartilage). The lack of physical simulation and pressure distribution optimization design makes it easy to cause pressure sores or pain in sensitive areas such as the bridge of the nose and temporal bone after wearing for a long time, and even cause secondary damage to fragile tissues after surgery.

[0027] In view of this, this application provides a personalized eyeglass frame design method based on 3D facial features, which can provide users with a precise frame solution with high adaptability, excellent comfort and aesthetic function.

[0028] The optional embodiments of this application are described in detail below with reference to the accompanying drawings.

[0029] The embodiment provided in this application is an embodiment of a personalized eyeglass frame design method based on 3D facial features.

[0030] The following is combined Figure 1 The embodiments of this application will be described in detail.

[0031] Figure 1 A flowchart illustrating a personalized eyeglass frame design method based on 3D facial features is shown, such as... Figure 1 As shown, it includes the following steps; Step S101: Obtain the user's facial scan data based on the scanning device.

[0032] Among them, facial scan data represents the complete set of scan data, which includes various types of information such as three-dimensional spatial location information, texture information, and color information.

[0033] The facial scan data includes at least facial 3D geometric data, which is used to describe the point cloud or mesh data of the three-dimensional shape of the facial surface, defining the undulations and contours of the face, and is used to provide data input for the High-Resolution Network (HRNet).

[0034] Step S102: Extract key feature points from facial 3D geometric data using the improved HRNet architecture.

[0035] Step S103: Take the key feature points as input and map them through a multilayer perceptron model to generate an initial frame parameter set containing 32 core parameters.

[0036] The 32 core parameters consist of four parts: frame geometry parameters, bridge of the nose area parameters, temple parameters, and overall design parameters.

[0037] Step S104: Construct a multi-dimensional constraint matrix consisting of geometric constraint submatrix, style constraint submatrix, manufacturing constraint submatrix, comfort constraint submatrix, and medical constraint submatrix.

[0038] Step S105: Establish a constrained optimization objective function, using the initial frame parameter set as the decision variable and the multi-dimensional constraint matrix as the constraint condition. Iteratively solve the optimal frame parameter set that minimizes the degree of violation of the constraints in each dimension through a sequential quadratic programming algorithm.

[0039] Step S106: Generate a 3D model of a personalized eyeglass frame based on the optimal frame parameter set.

[0040] In this embodiment, the problems of low adaptability of traditional eyeglass frame design, reliance on human experience, and difficulty in meeting special medical rehabilitation needs are solved. The entire process from patient facial data collection to personalized eyeglass frame model generation is automated and intelligent, and precise adaptation to the facial anatomical features of different subjects is achieved.

[0041] As a feasible embodiment, a high-precision 3D scanner, a time-of-flight (ToF) camera, or a multi-view photogrammetry unit is used in the process of acquiring facial scan data of the subject based on the scanning device.

[0042] In some embodiments, the obtained facial scan data includes at least facial 3D geometric data, and is subjected to denoising, smoothing and normalization by a data preprocessing layer to eliminate noise interference in the original point cloud.

[0043] As a specific embodiment, this application utilizes an improved HRNet architecture to extract key feature points from facial 3D geometric data. HRNet ensures the accuracy of feature localization by maintaining high-resolution representations and continuously fusing cross-resolution information.

[0044] In this embodiment, the key feature points are 22 feature points located at different positions on the face, and their specific anatomical distribution is as follows: (1) Eight feature points distributed in the facial contour area: left mandibular endpoint, right mandibular endpoint, left masseter point, right masseter point, left temporal point, right temporal point, mandibular projection point and facial tangent point.

[0045] (2) Six feature points distributed in the eye area: left outer corner point, right outer corner point, left inner corner point, right inner corner point, left eyebrow peak point and right eyebrow peak point.

[0046] (3) Four characteristic points distributed in the nasal region: nasal tip point, nasal root point, left nasal wing point and right nasal wing point.

[0047] (4) Four feature points distributed in the mouth area: left corner of mouth, right corner of mouth, center of upper lip and center of lower lip.

[0048] Among them, the feature points of the facial contour are used to establish the facial width benchmark and the path of the temple ear loops, the feature points of the eye area are used to define the geometric boundaries and tilt of the frame, the feature points of the nose area directly drive the calculation of the parameters of the bridge of the nose, and the feature points of the mouth area are used to analyze the vertical proportion of the face.

[0049] In this embodiment of the application, after the initial extraction is completed, the reliability score of each key feature point is calculated to evaluate the data quality.

[0050] For example, the reliability score is derived by weighted summation of the confidence score, consistency score, and spatial coherence score extracted from key feature points. For instance, if a feature point has a low confidence score due to occlusion, this application adjusts the original weights of each parameter in the optimal frame parameter set based on the reliability score, thereby reducing the influence of that point in subsequent calculations to ensure the robustness of the design.

[0051] In this embodiment, the reliability score is not a single metric, but rather a weighted sum of the confidence score, consistency score, and spatial coherence score from the feature point extraction process. The confidence score reflects the degree of certainty the deep learning model has regarding the point's location, the consistency score measures the stability of the point across consecutive frames or multiple viewpoints, and the spatial coherence score assesses the reasonableness of the geometric relationship between the point and its surrounding neighborhood points.

[0052] In some embodiments, this application uses key feature points as input and maps them through a multilayer perceptron model to generate an initial frame parameter set containing 32 core parameters.

[0053] Furthermore, the Multilayer Perceptron (MLP) was used as a regression predictor to establish a mapping from the coordinate space of 22 feature points to the space of 32 frame parameters. The 32 core parameters consist of four parts, including frame geometry parameters, bridge region parameters, temple parameters, and overall design parameters.

[0054] As a specific embodiment, the core parameters include at least the frame width, frame height, bridge width, nose bridge width, and temple length.

[0055] For example, the frame width is mapped according to the Euclidean distance between the left and right mandibular endpoints to ensure that the overall lateral dimensions of the frame precisely match the patient's face width.

[0056] For example, the height of the glasses frame is mapped based on the distance between the tip of the nose and the left brow peak, ensuring that the vertical dimension of the frame exactly covers the eye area.

[0057] For example, the width of the bridge, as a key part affecting wearing comfort, is mapped based on the distance between the left and right alar points and a preset tissue repair factor. This tissue repair factor can provide special compensation for postoperative or sensitive nasal areas, improving wearing safety.

[0058] For example, the width of the bridge of the nose is mapped based on the distance between the left and right alar points, combined with a width variation, to achieve precise control over the width of the nose pad support.

[0059] For example, the temple length is mapped based on the distance between the left temporal point and the projection point of the mandible, with an added length compensation value to ensure that the temples can bend naturally along the auricle and fit securely.

[0060] In some specific embodiments, the 32 core parameters consist of 12 frame geometry parameters, 6 bridge of the nose parameters, 8 temple parameters, and 6 overall design parameters.

[0061] For example, the 12 frame geometry parameters consist of 6 mapped basic dimension parameters (examples include frame width, frame height, inner diameter, top frame height, bottom frame height, and bridge width) and 6 shape parameters (examples include frame tilt angle, frame radius of curvature, top frame radius of curvature, bottom frame radius of curvature, frame thickness, and bridge radius of curvature). The parameters and their corresponding mapping formulas are as follows: Frame width: ; in, Indicates the width of the frame. This represents the first proportionality coefficient, and distance represents the Euclidean distance calculation function. Indicates the left mandibular endpoint. Indicates the right mandibular endpoint. This represents the second proportionality coefficient. This represents the coefficient of difference in width between the left and right sides.

[0062] Frame height: ;

[0063] in, Indicates the height of the frame. This represents the third proportionality coefficient, and distance represents the Euclidean distance calculation function. Indicates the tip of the nose. This indicates the left brow peak or brow arch point. This represents the fourth proportionality coefficient. This indicates the vertical height asymmetry coefficient.

[0064] Inner diameter: ; in, Indicates the inner diameter. This represents the fifth proportionality coefficient. This represents the function for calculating Euclidean distance. Represents the inner corner point of the left eye, This indicates the inner corner of the right eye.

[0065] Top frame height: ;

[0066] in, Indicates the height of the top frame. This represents the sixth proportionality coefficient. This represents the function for calculating Euclidean distance. Indicates the left eyebrow peak. It indicates the tip of the nose.

[0067] Bottom frame height: ; in, Indicates the height of the bottom frame. This represents the seventh proportionality coefficient. This represents the function for calculating Euclidean distance. Indicates the tip of the nose. This indicates the left corner of the mouth.

[0068] Bridge width: ; in, Indicates the width of the bridge section. This represents the eighth proportionality coefficient. This represents the function for calculating Euclidean distance. Indicates the left nasal ala point. Indicates the right nasal ala point. This represents the ninth proportionality coefficient. This indicates a tissue repair factor.

[0069] Frame tilt angle: ; in, Indicates the tilt angle of the picture frame. This represents the angle calculation function. Represents the inner corner point of the left eye, Represents the outer corner point of the left eye, Indicates the inner corner of the right eye. Indicates the outer corner of the right eye. This indicates the tilt correction amount.

[0070] Frame curvature radius: ; in, Indicates the radius of curvature of the picture frame. This represents the function for calculating curvature. Indicates the left mandibular endpoint. Indicates the left masseter muscle point. This indicates the right mandibular endpoint.

[0071] upper frame radius of curvature: ; in, This indicates the radius of curvature of the upper frame. This represents the function for calculating curvature. Indicates the left eyebrow peak. Indicates the left temporal point. This indicates the peak of the right eyebrow.

[0072] Radius of curvature of the lower frame: ; in, Indicates the radius of curvature of the lower frame. This represents the function for calculating curvature. Indicates the left mandibular endpoint. Indicates the left corner of the mouth. This indicates the right mandibular endpoint.

[0073] Frame thickness: ; in, Indicates the thickness of the frame. Indicates the base thickness of the frame. This indicates the thickness of the repair or the thickness of the tissue loss compensation. This represents the asymmetric factor.

[0074] Bridge curvature radius: ; in, Indicates the radius of curvature of the bridge section. This represents the function for calculating curvature. Indicates the root of the nose. Indicates the tip of the nose. This represents the feature data of the bridge of the nose.

[0075] For example, the parameters for the six nasal bridge regions and their corresponding mapping formulas are as follows: Nasal bridge width: ; in, Indicates the width of the bridge of the nose. This represents the function for calculating Euclidean distance. Indicates the left nasal ala point. Indicates the right nasal ala point. This represents the coefficient of difference in width between the left and right sides.

[0076] Nose pad position: ; in, Indicates the position of the nose pads. This represents the function for calculating Euclidean distance. Indicates the root of the nose. Indicates the tip of the nose. This indicates the positional scaling factor.

[0077] Nose bridge angle: ; in, Indicates the angle of the nose pads. This represents the angle calculation function. Indicates the root of the nose. Indicates the tip of the nose. Indicates the reference horizontal plane. This indicates the asymmetric correction amount.

[0078] Nasal pad curvature radius: ; in, Indicates the radius of curvature of the nose bridge. This represents the function for calculating curvature. Indicates the left nasal ala point. Indicates the tip of the nose. This indicates the right nasal ala point.

[0079] Nose bridge height: ; in, Indicates the height of the bridge of the nose. This represents the distance calculation function. Indicates the tip of the nose. Indicates the plan view of the bridge section. This indicates the height compensation value.

[0080] Nasal bridge tilt angle: ; in, Indicates the angle of inclination of the bridge of the nose. This represents the angle calculation function. Indicates the root of the nose. Indicates the tip of the nose. Indicates the reference vertical plane, This indicates the amount of tilt correction.

[0081] For example, the eight temple parameters and their corresponding mapping formulas are as follows: Temple length: ; in, Indicates the length of the temples. This represents the function for calculating Euclidean distance. Indicates the left temporal point. Indicates the projection point of the mandible. This represents the length compensation value.

[0082] Temple angle: ; in, Indicates the angle of the temples. This represents the function for calculating the temple angle. Indicates the left temporal point. Indicates the left mandibular endpoint. This indicates the point of tangency to the surface or the point of the surface profile.

[0083] Location of the temple bending point: ; in, Indicates the location of the bending point of the temple. Indicates the length of the temples. This represents the asymmetric factor.

[0084] Temple bending angle: ; in, Indicates the angle of bending of the temples. This indicates the amount of bending adjustment.

[0085] Temple curvature radius: ; in, Indicates the radius of curvature of the temple. This represents the function for calculating curvature. This represents path feature data from the temporal region to behind the ear.

[0086] Temple thickness: ; in, Indicates the thickness of the temples. Indicates the thickness of the frame. This indicates the thickness scaling factor.

[0087] Temple width: ; in, Indicates the width of the temples. This indicates the initial reference value for the temple width. Indicates the width scaling factor. This indicates the amount of asymmetric compensation.

[0088] Temple tip angle: ; in, Indicates the angle of the temple tips. This indicates the amount of correction for the tip angle.

[0089] For example, the six overall design parameters and their corresponding mapping formulas are as follows: Frame style factor: ; in, Indicates the frame style factor. This represents the function for calculating style scores. This represents a facial feature vector.

[0090] Bridge design type: ; in, Indicates the bridge design type. This represents the function for selecting the bridge type. This represents the geometric feature data of the bridge section.

[0091] Temple design types: ; in, Indicates the type of temple design. This represents the function for selecting the temple type. This represents the characteristic data of the temples. This indicates a patient's specific medical or functional needs.

[0092] Frame shape factor: ; in, Indicates the frame shape factor. This represents the function for calculating the shape factor. This represents eyebrow feature data. This represents data on facial contour points or tangent points.

[0093] Design complexity coefficient: ; in, This represents the design complexity coefficient. This represents the coefficient indicating the difference in width between the left and right sides. This represents the vertical height asymmetry coefficient.

[0094] Manufacturing feasibility coefficient: ; in, Indicates the manufacturing feasibility coefficient. This represents the manufacturing feasibility assessment function. This represents a parameter vector composed of frame design parameters.

[0095] In some embodiments, for the generated initial parameters, this application establishes a constrained optimization objective function, using the initial frame parameter set as the decision variable and the multi-dimensional constraint matrix as the constraint condition, and iteratively solves the optimal frame parameter set that minimizes the degree of violation of the constraints in each dimension through a sequential quadratic programming (SQ) algorithm.

[0096] The sequential quadratic programming algorithm is based on the following basic process: 1. Initialize the parameter vector ; 2. Construct the Lagrange function: ; 3. Iterative solution: a. Calculate the gradient: ; b. Construct a quadratic programming subproblem; c. Solve the subproblems to obtain the search direction; d-line search determines the step size; e. Update parameters: ; 4. Check convergence conditions: ; 5. Output the optimal solution .

[0097] in, Indicates the independent variable / decision variable. Represents the initial parameter vector. Describe the objective function. Represents the constraint function. Represents the Lagrange multipliers. Represent the Lagrange function, This represents the gradient of the Lagrange function. Indicates the number of iterations / steps. Indicates the search direction. Indicates step size, This represents the updated parameter vector. Indicates the convergence threshold / precision. This represents the optimal solution.

[0098] In some embodiments, this application constructs a five-layer constraint matrix system to ensure that the design meets multiple constraints. The overall architecture of the constraint matrix is ​​as follows: ;

[0099] in, Indicates a general constraint. Represents the geometric constraint matrix. Represents the style constraint matrix. Represents the manufacturing constraint matrix. Represents the comfort constraint matrix. This represents the medical constraint matrix.

[0100] For example, each submatrix contains corresponding constraints, as shown below.

[0101] Geometric constraint matrix: C_geometry=[ [c11,c12,...,c1n], [c21,c22,...,c2n], ... [cm1,cm2,...,cmn] ]

[0102] Where cij represents the correlation strength between the i-th parameter and the j-th geometric constraint.

[0103] Style constraint matrix: C_style=[ [s11,s12,...,s1k], [s21,s22,...,s2k], ... [sm1,sm2,...,smk] ]

[0104] Where sij represents the compatibility score between the i-th parameter and the j-th style.

[0105] Creating the constraint matrix: C_manufacture=[ [m11,m12,...,m1p], [m21,m22,...,m2p], ... [mp1,mp2,...,mpp] ] Where mij represents the difficulty coefficient of the i-th parameter for the j-th manufacturing process.

[0106] Comfort constraint matrix: C_comfort=[ [co11,co12,...,co1q], [co21,co22,...,co2q], ... [com1,com2,...,comq] ] Where coij represents the influence coefficient of the i-th parameter on the j-th comfort index.

[0107] Medical constraint matrix (C_medical): C_medical=[ [med11,med12,...,med1r], [med21,med22,...,med2r], ... [medm1,medm2,...,medmr] ] Where medij represents the constraint strength of the i-th parameter on the j-th medical indicator.

[0108] In some embodiments, during the process of using a sequential quadratic programming algorithm to iteratively solve for multi-objective optimization, situations may arise where different constraints conflict with each other.

[0109] To address this issue, in some embodiments, this application employs a three-level priority mechanism to handle constraint conflicts during the iterative solution process.

[0110] For example, this application sets geometric feasibility, structural strength, and manufacturing feasibility as the first priority hard constraints that cannot be violated; sets aesthetic requirements, style consistency, and comfort as the second priority soft constraints that can be moderately violated; and sets weight minimization, cost minimization, and production efficiency as the third priority objective constraints.

[0111] Based on this, when a constraint conflict occurs, parameters are adjusted first to satisfy the first priority hard constraint, and the second priority soft constraint is handled by minimizing the degree of violation, provided that the first priority hard constraint is satisfied.

[0112] In view of the above embodiments, this application further designs a conflict resolution algorithm for handling constraint conflicts, the core logic of which is as follows: First, the algorithm identifies all conflicting constraints and sorts them according to priority (hard constraints > soft constraints > target constraints). Then, it processes each conflict sequentially from highest to lowest priority: for hard constraint conflicts, the constraints must be fully satisfied by adjusting relevant parameters; for soft constraint conflicts, the algorithm first assesses whether the current violation is acceptable; if acceptable, it ignores the conflict; otherwise, it adjusts parameters to minimize the violation; for target constraints, it performs optimization to pursue the optimal objective. After processing all conflicts, the algorithm verifies the validity of the current solution. If the verification passes, the solution is returned; otherwise, a backtracking mechanism is triggered to solve the problem again.

[0113] This application employs a tiered conflict resolution logic, enabling the implementation of the solution to efficiently find the optimal balance between feasibility and comfort within a complex design space.

[0114] Regarding the above embodiments, this application can dynamically adjust the constraint weights according to different scenarios.

[0115] For example, in a typical customization scenario, the constraint weights are configured as follows: ; ; ; ; ; For example, in the case of severe facial asymmetry, the constraint weights are configured as follows: ; ; ; ; ; For example, in the early postoperative repair stage, the constraint weights are configured as follows: ; ; ; ; ; In some embodiments, this application constructs a finite element model of the eyeglass frame based on an optimal set of frame parameters, and reconstructs a facial finite element model including the biomechanical properties of skin, subcutaneous fat, and cartilage tissue using facial scan data. The elastic modulus of skin tissue is defined as 0.42-0.85 MPa, and the density of cartilage tissue is defined as 1.8 g / cm³. Subsequently, this application applies a self-weight load, facial muscle force based on electromyography data, and a periodic load based on blink frequency to the finite element model of the eyeglass frame. During the calculation, this application calculates the natural frequency of the eyeglass frame structure represented by the finite element model, and if the natural frequency is within the blink frequency range (e.g., 4Hz to 8Hz), adjusts the parameters of the finite element model of the eyeglass frame (e.g., local stiffness or mass distribution) until the natural frequency avoids the blink frequency range to prevent resonance during wear. After completing the dynamic analysis, this application determines the von Mises stress and contact pressure distribution in the contact area between the finite element model of the eyeglass frame and the finite element model of the face. Based on the von Mises stress and contact pressure distribution, it identifies the intra-domain pressure in the nasal bridge region, temporal bone region, and mandibular border region, and determines whether the intra-domain pressure exceeds a preset pain threshold. If there is intra-domain pressure exceeding the preset pain threshold (e.g., exceeding 0.15 MPa in the nasal bridge region), geometric optimization is performed on the corresponding region in the eyeglass frame finite element model, increasing the corner radius and / or adjusting the wall thickness distribution.

[0116] Based on this, this application generates a 3D model of a personalized eyeglass frame according to the optimal frame parameter set and outputs it as the result. Furthermore, the geometrically optimized finite element model of the frame is converted into a 3D printing format, and slicing optimization is performed according to additive manufacturing technology to generate a support structure, thereby completing the manufacturing preparation for the subject's custom-made eyeglass frame.

[0117] In the above embodiments, finite element analysis optimization is the core step in the personalized eyeglass frame design process, and it plays the role of performing in-depth simulation of the frame structure strength, wearing comfort and dynamic stability.

[0118] In some embodiments, this application predefines various material models and biomechanical parameters for different application scenarios.

[0119] As one feasible embodiment, the frame material model includes titanium alloy (Ti-6Al-4V), medical stainless steel (316L), and carbon fiber reinforced polymer (CFRP).

[0120] Among them, the density of titanium alloy materials elastic modulus Poisson's ratio Yield strength Fatigue limit coefficient of thermal expansion Density of medical-grade stainless steel elastic modulus Poisson's ratio Yield strength Fatigue limit coefficient of thermal expansion Density of carbon fiber reinforced polymer Fiber-direction elastic modulus Poisson's ratio Yield strength Fatigue limit coefficient of thermal expansion .

[0121] As a specific embodiment, this application establishes a biomechanical material parameter system for the facial tissue of the subject to support subsequent pressure sensitivity simulation.

[0122] Specifically, the density of skin tissue Its elastic modulus exhibits strain dependence, ranging from Poisson's ratio viscoelastic parameters The density of subcutaneous fat elastic modulus Poisson's ratio viscoelastic parameters Density of cartilage tissue The elastic modulus is strain dependent Poisson's ratio Permeability coefficient .

[0123] In some embodiments, this application constructs a finite element model of the eyeglass frame based on an optimal frame parameter set, generates geometric entities using parametric modeling techniques, optimizes surface smoothness through a hybrid deformation model, and then generates a high-quality mesh to ensure element size. .

[0124] Simultaneously, this application reconstructs a facial finite element model incorporating the biomechanical properties of skin, subcutaneous fat, and cartilage tissue using facial scan data. A hexahedral-dominant mesh generation method is employed, with approximately 500,000 elements. Mesh refinement techniques are applied to the predicted contact areas to improve computational accuracy. Based on this, this application establishes a contact surface model by identifying the contact area between the eyeglass frame and the face, employing either the penalty function method or the Lagrange multiplier method to set the contact algorithm, and setting the friction coefficient between the skin and metal to [value missing]. .

[0125] To simulate realistic wearing conditions, this application sets complex boundary conditions and load environments. Regarding displacement constraints, the bottom of the face is set to full constraint to simulate head fixation, symmetry constraints are applied to the frame's symmetry planes, and contact constraints are set for the contact surfaces. During load application, this application applies self-weight load, facial muscle forces based on electromyography data, and periodic loads based on blink frequency to the finite element model of the frame. The self-weight load... It exists in the form of distributed volumetric forces, and the formula is: Facial muscle strength based on electromyography (EMG) data The formula is Periodic load based on blink frequency Frequency set to The calculation formula is: .

[0126] As a feasible embodiment, this application performs dynamic response analysis before formally evaluating static pressure sensitivity. This application calculates the natural frequencies of the eyeglass frame structure represented by the finite element model of the frame, and the solution equation is as follows: ,in Here is the stiffness matrix. For the quality matrix, This refers to the natural frequency. If the natural frequency falls within the blink frequency range, that is, within... In the high-risk resonance range, this application adjusts the parameters of the finite element model of the eyeglass frame, for example by changing the support structure, adding stiffeners or adjusting the load path to change the structural stiffness distribution until the natural frequency avoids the blink frequency range, so as to minimize the vibration amplitude and increase the structural damping ratio.

[0127] Subsequently, static strength and pressure analysis was performed to determine the von Mises stress and contact pressure distribution in the contact area between the finite element model of the eyeglass frame and the finite element model of the face. The von Mises stress was calculated using the following formula: ; Among them, contact pressure Then the normal contact force Divide by actual contact area Obtained. During this process, this application will also calculate a pressure distribution uniformity index, using the formula: ,in For pressure standard deviation, This represents the average pressure.

[0128] As a specific embodiment, this application identifies the intra-regional pressure in the nasal bridge region, temporal bone region, and mandibular border region based on the von Mises stress and contact pressure distribution, and determines whether the intra-regional pressure exceeds a preset pain threshold. For the nasal bridge region (nasal pad contact area), the preset pain threshold is... Damage threshold For the temporal bone region (upper edge of the temple), the pain threshold... Damage threshold For the mandibular region (lower edge of the temples), the pain threshold... Damage threshold .

[0129] If pressure exceeds a preset pain threshold within the target area, this application will initiate a pressure optimization algorithm, constructing an objective function that minimizes the weighted squared difference between the actual pressure and the target pressure. At this point, this application performs geometric optimization on the region corresponding to the pressure within the target area in the finite element model of the eyeglass frame. Specific geometric optimizations include increasing the corner radius and / or adjusting the wall thickness distribution. The calculation logic for increasing the corner radius is as follows: Adjusting the wall thickness distribution is achieved through... After completing the geometric optimization, this application re-evaluates the pressure distribution and checks the convergence conditions until the optimal design scheme that meets the comfort requirements is output.

[0130] Furthermore, this application also possesses multi-material integrated optimization capabilities, prioritizing high-elastic-modulus materials such as titanium alloys or CFRP for areas requiring strength, while prioritizing low-density materials for areas requiring lightweighting. Through functionally graded material design, the elastic modulus is... Can be used depending on location According to the gradient distribution function Dynamically changing, the formula is: Regarding environmental adaptability, this application performs thermal expansion analysis and calculates thermal strain. and thermal stress Furthermore, the subject's skin temperature perception was assessed using thermal comfort indices. To ensure product reliability, this application predicts fatigue life based on SN curve theory and Miner's rule, optimizing fatigue strength by reducing stress concentration and lowering the load spectrum.

[0131] Furthermore, in order to adapt to the specific clinical conditions of different patients, the method provided in this application also has adaptive adjustment capabilities.

[0132] As a feasible embodiment, this application can intelligently identify the facial features or clinical status of the subject, and the weight adjustment logic is configured to have environmental awareness capabilities.

[0133] For example, when the system detects severe facial asymmetry, it automatically increases the weight of the geometric constraint submatrix in the overall optimization objective function to prioritize ensuring that the frames can accurately fit the asymmetrical facial structure.

[0134] Specifically, when facial asymmetry is detected, the weight of the geometric constraint submatrix can be increased, for example, by setting it to 0.4.

[0135] For example, when the system detects that the subject is in the early postoperative recovery stage, it will increase the weights of the comfort constraint submatrix and the medical constraint submatrix, so that the optimization algorithm will focus more on reducing pressure and stimulation on the surgical area and prioritize the patient's recovery experience and wearing safety.

[0136] Specifically, when a subject is identified as being in the early postoperative recovery phase, the weights of the comfort constraint submatrix and the medical constraint submatrix can be increased to address special adaptation needs such as facial trauma. For example, the comfort constraint can be set to 0.3 and the medical constraint to 0.15.

[0137] This application achieves significant technological advancements through deep integration of hardware carrier, algorithm architecture, and mechanical analysis process. The technical effects of the aforementioned personalized eyeglass frame design method based on 3D facial features in practical applications are described below.

[0138] In some embodiments, this application demonstrates extremely strong and precise adaptation capabilities. This application acquires facial scan data of the subject using a scanning device, providing a geometric benchmark with micrometer-level precision for subsequent design. Utilizing an improved HRNet architecture, key feature points are extracted from the facial 3D geometric data. This model achieves sub-pixel-level localization of 22 feature points located at different positions on the face by maintaining high-resolution representation and continuously performing multi-scale feature fusion. This application further uses these key feature points as input and maps them through a multilayer perceptron model. This multilayer perceptron consists of an input layer, multiple hidden layers, and an output layer, and can non-linearly generate an initial frame parameter set containing 32 core parameters. Through this parameterization system, this application achieves digital reconstruction of the subject's complex anatomical morphology. In particular, by calculating the left-right width difference coefficient and the vertical height asymmetry coefficient, facial asymmetry features are quantified, thereby generating a frame structure that can automatically compensate for facial defects or tissue loss, solving the problem in medical rehabilitation scenarios where standard frames cannot adapt to facial trauma or postoperative anatomical changes.

[0139] As a feasible embodiment, this application achieves refined multi-dimensional constraint optimization. Through a mathematical optimization module deployed on a computing terminal, this application constructs a multi-dimensional constraint matrix consisting of geometric constraint sub-matrices, style constraint sub-matrices, manufacturing constraint sub-matrices, comfort constraint sub-matrices, and medical constraint sub-matrices, establishing five-dimensional evaluation criteria for the design scheme. This application establishes a constraint optimization objective function, using the initial frame parameter set as decision variables and the multi-dimensional constraint matrix as constraint conditions, aiming to find the parameter vector that minimizes the degree of violation. During the solution process, this application uses a sequential quadratic programming algorithm for iterative solution. This algorithm transforms the nonlinear constrained problem into a series of quadratic programming sub-problems, maximizing aesthetic consistency and wearing comfort while ensuring that hard constraints such as geometric feasibility, structural strength, and manufacturing feasibility are forcibly satisfied. This multi-criteria balancing mechanism ensures that the generated 3D model can still approach the global optimum under multiple boundary conditions.

[0140] As a specific embodiment, this application demonstrates advanced physical simulation capabilities. Based on an optimal set of frame parameters, this application constructs a finite element model of the eyeglass frame and reconstructs a facial finite element model incorporating the biomechanical properties of skin, subcutaneous fat, and cartilage tissue using facial scan data. During the simulation, this application considers the nonlinear properties of materials and complex contact nonlinearities, and applies self-weight loads, facial muscle forces based on electromyography data, and periodic loads based on blink frequency to the eyeglass frame finite element model. By determining the von Mises stress and contact pressure distribution in the contact area between the eyeglass frame finite element model and the facial finite element model, this application can proactively identify potential pain areas. By calculating the natural frequencies of the eyeglass frame structure characterized by the eyeglass frame finite element model, this application avoids the risk of dynamic resonance, ensuring excellent dynamic stability of the eyeglass frame during blinking or movement, completing a technological iteration from simple "geometric fit" to deep "functional comfort."

[0141] For example, this application possesses a high degree of intelligence. Integrated into an automated design software platform, it utilizes an improved HRNet model and a multilayer perceptron model deployed on the server side to achieve automatic feature point extraction and real-time parameter mapping. This fully automated pipeline eliminates the reliance on manual measurement and subjective experience in traditional frame design, significantly reducing the time from acquiring scan data to generating the optimized 3D model and improving design efficiency. Furthermore, the feature point quality assessment mechanism employed in this application—which calculates the reliability score of each key feature point and adjusts the original weights of parameters in the optimal frame parameter set accordingly—significantly enhances the algorithm's tolerance to noisy data, ensuring high consistency and repeatability of the design output.

[0142] As a specific embodiment, this application demonstrates significant clinical value. In the field of medical rehabilitation, this application increases the weight of the geometric constraint submatrix in response to the detection of facial asymmetry; or increases the weight of the comfort constraint submatrix and the medical constraint submatrix in response to the detection that the subject is in the early postoperative recovery stage. This scenario-adaptive parameter control strategy provides patients with facial trauma with a customized solution that can avoid sensitive areas and compensate for damaged tissue. By geometrically optimizing the pressure-corresponding region within the target domain in the finite element model of the eyeglass frame, this application ensures that the eyeglass frame can be worn for extended periods without causing skin damage, significantly improving the quality of life and social integration of patients in the recovery period.

[0143] This application also provides system embodiments that follow the above embodiments, for implementing the method steps of the above embodiments. The interpretation of the same names is the same as that of the above embodiments, and they have the same technical effects as those of the above embodiments, so they will not be repeated here.

[0144] like Figure 2 As shown, this application provides a personalized eyeglass frame design system 200 based on 3D facial features, comprising: The data acquisition module 201 is used to acquire the user's facial scan data based on the scanning device. The facial scan data includes at least 3D facial geometric data.

[0145] Feature extraction module 202 is used to extract key feature points from facial 3D geometry data using an improved HRNet architecture.

[0146] The parameter generation module 203 is used to take key feature points as input, map them through a multilayer perceptron model, and generate an initial frame parameter set containing 32 core parameters. The 32 core parameters consist of four parts: frame geometry parameters, bridge of the nose region parameters, temple parameters, and overall design parameters.

[0147] The constraint optimization module 204 is used to construct a multi-dimensional constraint matrix consisting of geometric constraint submatrices, style constraint submatrices, manufacturing constraint submatrices, comfort constraint submatrices, and medical constraint submatrices.

[0148] The finite element optimization module 205 is used to establish a constrained optimization objective function. It uses the initial frame parameter set as the decision variable and the multi-dimensional constraint matrix as the constraint condition. It iteratively solves the optimal frame parameter set that minimizes the degree of violation of the constraints in each dimension through a sequential quadratic programming algorithm.

[0149] Manufacturing implementation module 206 is used to generate a 3D model of a personalized frame based on the optimal frame parameter set.

[0150] Regarding the system in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0151] Although the operations are described in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order or serial order shown, or requiring all of the operations shown to obtain the desired result. In certain environments, multitasking and parallel processing may be advantageous.

[0152] The methods and systems of this application can be implemented using standard programming techniques, utilizing rule-based logic or other logic to implement various method steps. It should also be noted that the terms "system" and "module" as used herein and in the claims are intended to include implementations using one or more lines of software code and / or hardware implementations and / or devices for receiving input.

[0153] Any step, operation, or procedure described herein may be performed or implemented using one or more hardware or software modules, either alone or in combination with other devices. In one embodiment, the software module is implemented using a computer program product comprising a computer-readable medium containing computer program code, which is executable by a computer processor to perform any or all of the described steps, operations, or procedures.

[0154] The foregoing description of implementations of this application has been provided for illustrative and descriptive purposes. The foregoing description is not exhaustive and is not intended to limit this application to the exact forms disclosed. Various modifications and variations may exist in accordance with the foregoing teachings, or may arise from practice of this application. These embodiments were chosen and described to illustrate the principles of this application and its practical application, enabling those skilled in the art to utilize this application in various implementations and modifications to suit the specific purpose of the concept.

[0155] Regarding the system in the above embodiments, the specific ways in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0156] It can be further understood that, unless otherwise specified, "connection" includes both direct connections where no other components exist between the two parties and indirect connections where other components exist between them.

[0157] It is further understood that although the operations are described in a specific order in the accompanying drawings in the embodiments of this application, this should not be construed as requiring these operations to be performed in the specific order or serial order shown, or requiring all the operations shown to be performed to obtain the desired result. In certain environments, multitasking and parallel processing may be advantageous.

[0158] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the field of this application that are not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0159] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

[0160] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for designing personalized eyeglass frames based on 3D facial features, characterized in that, include: The user's facial scan data is acquired using a scanning device; wherein the facial scan data includes at least facial 3D geometric data; Key feature points are extracted from the facial 3D geometry data using an improved HRNet architecture; The key feature points are used as input and mapped through a multilayer perceptron model to generate an initial frame parameter set containing 32 core parameters. The 32 core parameters consist of four parts: frame geometry parameters, bridge of the nose parameters, temple parameters, and overall design parameters. Construct a multi-dimensional constraint matrix consisting of geometric constraint submatrices, style constraint submatrices, manufacturing constraint submatrices, comfort constraint submatrices, and medical constraint submatrices; Establish a constrained optimization objective function, using the initial frame parameter set as the decision variable and the multi-dimensional constraint matrix as the constraint condition, and iteratively solve the optimal frame parameter set that minimizes the degree of violation of each dimension constraint condition through a sequential quadratic programming algorithm; A 3D model of a personalized eyeglass frame is generated based on the optimal frame parameter set.

2. The personalized eyeglass frame design method based on 3D facial features according to claim 1, characterized in that, The key feature points are 22 feature points located at different positions on the face, including: The left mandibular endpoint, right mandibular endpoint, left masseter point, right masseter point, left temporal point, right temporal point, mandibular projection point, and facial tangent point are distributed in the facial contour area; The outer corner of the left eye, the outer corner of the right eye, the inner corner of the left eye, the inner corner of the right eye, the left eyebrow peak, and the right eyebrow peak are distributed in the eye area. The nasal tip, nasal root, left nasal alar, and right nasal alar are distributed in the nasal region; and The left corner of the mouth, the right corner of the mouth, the center of the upper lip, and the center of the lower lip are distributed in the mouth area.

3. The personalized eyeglass frame design method based on 3D facial features according to claim 2, characterized in that, The core parameters include at least the frame width, frame height, bridge width, nose bridge width, and temple length, which are obtained by mapping as follows: The width of the eyeglass frame is mapped based on the Euclidean distance between the left mandibular endpoint and the right mandibular endpoint; The height of the glasses frame is mapped based on the distance between the tip of the nose and the left brow peak. The bridge width is mapped based on the distance between the left and right alar points combined with a preset tissue repair factor; The width of the bridge of the nose is mapped by combining the distance between the left and right alar points with the width change. The temple length is mapped based on the distance between the left temporal point and the mandibular projection point, combined with a length compensation value.

4. The personalized eyeglass frame design method based on 3D facial features according to claim 1, characterized in that, The method further includes: Calculate the reliability score for each of the key feature points; wherein the reliability score is obtained by weighted summation of the confidence score, consistency score and spatial coherence score extracted from the key feature points; The original weights of each parameter in the optimal frame parameter set are adjusted based on the reliability score.

5. The personalized eyeglass frame design method based on 3D facial features according to claim 1, characterized in that, During the iterative solution process, a three-level priority mechanism is used to handle constraint conflicts. The priority rules between different constraints are as follows: Geometric feasibility, structural strength, and manufacturing feasibility are set as the first-priority hard constraints that cannot be violated; Aesthetic requirements, stylistic consistency, and comfort are set as the second priority soft constraints that can be moderately violated; Minimize weight, minimize cost, and improve production efficiency as the third priority objective constraints; When a constraint conflict occurs, parameters are adjusted first to satisfy the first priority hard constraint, and the second priority soft constraint is handled by minimizing the degree of violation, provided that the first priority hard constraint is satisfied.

6. The personalized eyeglass frame design method based on 3D facial features according to claim 1, characterized in that, The method further includes: A finite element model of the eyeglass frame is constructed based on the optimal frame parameter set, and a finite element model of the face containing the biomechanical properties of skin, subcutaneous fat and cartilage tissue is reconstructed by combining the facial scan data. The frame finite element model is subjected to a self-weight load, facial muscle force based on electromyography data, and a periodic load based on blink frequency. Determine the von Mises stress and contact pressure distribution in the contact area between the finite element model of the eyeglass frame and the finite element model of the face; Based on the von Mises stress and the contact pressure distribution, the intra-domain pressure of the nasal bridge region, temporal bone region and mandibular border region is identified, and it is determined whether the intra-domain pressure of each region exceeds the preset pain threshold. If there is pressure within the target domain exceeding the preset pain threshold, then geometric optimization is performed on the region corresponding to the pressure within the target domain in the finite element model of the eyeglass frame. The geometry optimization includes increasing the fillet radius and / or adjusting the wall thickness distribution.

7. The personalized eyeglass frame design method based on 3D facial features according to claim 6, characterized in that, Before determining the von Mises stress and the contact pressure distribution, the method further includes: Calculate the natural frequencies of the eyeglass frame structure as represented by the finite element model of the frame; If the natural frequency is within the blink frequency range, then adjust the parameters of the finite element model of the eyeglass frame until the natural frequency avoids the blink frequency range.

8. The personalized eyeglass frame design method based on 3D facial features according to claim 6, characterized in that, The process of generating a 3D model of a personalized eyeglass frame based on the optimal frame parameter set includes: The geometrically optimized finite element model of the eyeglass frame is converted into a 3D printing format, and the slicing optimization is performed according to the additive manufacturing process to generate the support structure.

9. The personalized eyeglass frame design method based on 3D facial features according to claim 1, characterized in that, The method further includes: In response to the detection of facial asymmetry, the weights of the geometric constraint submatrix are increased; In response to the identification that the user is in the early postoperative recovery period, the weights of the comfort constraint submatrix and the medical constraint submatrix are increased.

10. A personalized eyeglass frame design system based on 3D facial features, characterized in that, include: A data acquisition module is used to acquire facial scan data of a user based on a scanning device; wherein, the facial scan data includes at least facial 3D geometric data; A feature extraction module is used to extract key feature points from the facial 3D geometry data using an improved HRNet architecture; The parameter generation module is used to take the key feature points as input, map them through a multilayer perceptron model, and generate an initial frame parameter set containing 32 core parameters; wherein, the 32 core parameters consist of four parts including frame geometry parameters, bridge of nose region parameters, temple parameters, and overall design parameters; The constraint optimization module is used to construct a multi-dimensional constraint matrix consisting of geometric constraint submatrices, style constraint submatrices, manufacturing constraint submatrices, comfort constraint submatrices, and medical constraint submatrices. The finite element optimization module is used to establish a constrained optimization objective function, using the initial frame parameter set as the decision variable and the multi-dimensional constraint matrix as the constraint condition, and iteratively solving the optimal frame parameter set that minimizes the degree of violation of the constraints in each dimension through a sequential quadratic programming algorithm. The manufacturing implementation module is used to generate a 3D model of a personalized eyeglass frame based on the optimal frame parameter set.

Images