Frame adjustment system

Abstract

A computer-implemented method for fitting eyeglass frames to a user to provide a customized eyeglass frame, the method comprising the steps of receiving input data including three-dimensional coordinate data representative of a user's head; identifying a plurality of landmark locations in the input data, the landmark locations corresponding to three-dimensional locations of facial features; determining a facial measurement set based on the plurality of landmark locations by calculating at least one measurement associated with the at least one landmark location; retrieving a frame measurement set representative of a basic eyeglass frame from a database of frame measurements; comparing the facial measurement set to the frame measurement set; adjusting at least one frame measurement of the frame measurement set based on the comparison; and outputting a data file including the adjusted frame measurement set, the frame measurement set including lens height measurements, and the adjusting step including adjusting the lens height measurements from an initial value to an adjusted value.

Description

Technical Field

[0001] The present invention relates to a frame adjustment system.

Background Art

[0002] An eyeglass frame has the main role of holding a pair of lenses of the required power and type for vision improvement in front of the wearer's both eyes at the correct position and angle.

[0003] A well-fitted pair of glasses or eyeglasses needs to meet several major criteria to function properly when positioning the lenses correctly with respect to the wearer's pupils to provide vision improvement to the wearer.

[0004] 1. Optimization of optical performance The optical centers of the lenses should be properly positioned in front of both eyes by the frame so as to provide the wearer with optimal vision correction.

[0005] The optical centers of the lenses of the correct power and type should be placed horizontally (on the x-axis of the Cartesian coordinate system via accurate interpupillary distance measurement) and vertically (on the y-axis of the Cartesian coordinate system via accurate height measurement) in front of the pupils.

[0006] The lenses should be positioned vertically in alignment with the vertical plane. That is, when the pupils correspond to the optical centers of the lenses, no front tilt should be provided. The addition of lens tilt changes the spherical power of the lenses and causes an unwanted cylindrical component to appear. For low-power lenses, the power effect is slight, but it can be extremely obvious for high-power lenses.

[0007] The selected frame should be useful for specific types of prescriptions including high-index single-focus lenses, aspheric lenses, progressive lenses, etc.

[0008] 2. Maximization of the visual field The primary purpose of the lens is to maximize the wearer's field of vision. Since the field of vision increases and magnification decreases as the orthogonal lens approaches both eyes, the frame should keep the lens as close to both eyes as possible and minimize the intervertebral distance, taking into account the influence of eyelashes, cheekbones, and other obstructions. The wearer's eyelashes should be slightly away from the back surface of the lens.

[0009] The selected frame should function with respect to the wearer's unique and specific facial features, including bridge height, bridge protrusion, nose size and shape, head size, and asymmetry.

[0010] 3. Maintaining consistency The frame should be selected for its ability to maintain its alignment, and should minimize movement to prevent it from slipping off the nose.

[0011] This not only irritates the wearer but also has serious optical side effects. These include: 1. blurred distance vision as a result of increased effective lens power; 2. reduced field of view; 3. increased magnification of objects seen by the wearer; and 4. increased apparent size of the wearer's eyes to the observer (for orthogonal lenses, a decrease in vertex distance makes the wearer's eyes appear less magnified to the observer).

[0012] Providing comfort When worn, eyeglass frames and lenses should be comfortable for the wearer. The weight of the frame should be evenly distributed across the bridge surface and where it rests on the ears. There should be no points where the frame feels too tight.

[0013] In order to provide the wearer with a complete set of eyeglasses that provide the necessary vision correction, eyeglass frames may be fabricated or adjusted to fit the wearer's face based on measurements taken of the wearer's head.

[0014] Currently, there are four main methods for obtaining these measurements, as outlined below.

[0015] The first, and most commonly used, method is trial and error—the wearer tries on frames, with or without the assistance of an eye examiner or salesperson, using frame measurements typically written on the inside of the temples for guidance, until they find a satisfactory fit. If these measurements are not directly related to the fit of the frame when worn, this approach, coupled with a generally insufficient understanding of optical fit, generally leads to unsatisfactory results.

[0016] The second method requires the optician to manually obtain measurements of the wearer's head using a ruler by marking various points on the wearer's face with a pen and then measuring the distances between these marked points. This method involves several inherent inaccuracies, such as human error when making the initial marks and when taking the actual measurements, patient or measurer movement during the measurement process, and facial asymmetry. In addition, the width of the pen becomes important when dealing with small measurements related to facial features, thus introducing further inaccuracies into the measurement process.

[0017] The third method requires obtaining measurements using electronic measuring tools, for example, a pupillary meter to measure the distance between the wearer's pupils. While electronic measuring tools can mitigate some of the problems associated with manual measurement, some system inaccuracies that occur even with highly functional electronic devices still exist. For example, electronic measuring tools cannot eliminate inaccuracies associated with incorrect placement of the measuring tool by the person taking the measurement.

[0018] The fourth method requires the use of a still image or photograph of the wearer's head and the acquisition of measurements using a datum or scaling device. This is sometimes referred to as the online "credit card" technique, in which one edge of a credit card is placed at the center of the wearer's nose and the point at the end of the credit card is measured. The problem with this method is that it is difficult to ensure that the credit card is positioned in the correct vertical plane when superimposed on an image of the wearer's face, leading to the acquisition of inaccurate measurements. Furthermore, inconsistencies in lens curvature, along with camera curvature, introduce small distortions into the image that result in inaccurate measurements.

[0019] Therefore, it is clear that there are many inherent problems associated with providing wearers with eyeglasses that are properly adjusted to their heads so that the eyeglasses hold the lenses in the correct position relative to the wearer's pupils.

[0020] In this regard, there are several key parameters that are important to measure accurately in order to enable proper adjustment of eyeglass frames. These parameters include interpupillary distance, vertex distance, forward tilt, and lens segment height.

[0021] Interpupillary distance is important because if the wearer is not looking at the optical center of the lens, the resulting incorrect prescription and induced prism refractive power will affect visual clarity and comfort. Regarding the forward tilt, lenses within a frame are usually not positioned vertically to the vertical plane, thus inducing changes in the lens prescription. Regarding the vertex distance, when the vertex distance of the lens is changed, the amount of light reaching the front of the cornea also changes, affecting the lens's ability to provide the wearer with the required visual correction. Finally, if the height measurement is inaccurate, the wearer may be looking with the wrong prescription in their primary gaze rather than their auxiliary gaze.

[0022] Thus, in order to properly fit lenses to an individual's pupil, it is necessary to provide a method for adjusting the entire eyeglasses set to fit the individual's head. [Overview of the project] [Means for solving the problem]

[0023] According to the first embodiment, a computer-based method is provided that adjusts the eyeglass frame to suit the user and provides a customized eyeglass frame, comprising the steps of: receiving input data that includes three-dimensional coordinate data representing the user's head; identifying a plurality of landmark locations in the input data, wherein the landmark locations correspond to the three-dimensional positions of facial features; determining a set of facial measurement values ​​based on the plurality of landmark locations by calculating at least one measurement value associated with at least one landmark location using the landmark locations; searching a database of frame measurement values ​​for a set of frame measurement values ​​representing a basic eyeglass frame; comparing the set of facial measurement values ​​with the set of frame measurement values; adjusting at least one frame measurement value in the set of frame measurement values ​​based on the comparison; and outputting a data file that includes the adjusted set of frame measurement values, wherein the set of frame measurement values ​​includes lens height measurement values, and the adjustment step includes adjusting the lens height measurement value from an initial value to an adjusted value.

[0024] In this context, the measurements may include angles, such as the angle at which the nose protrudes from the face, or distances between two points, such as the distance between pupils. In these examples, the nose and pupils are landmarks corresponding to the relevant angle and distance measurements, respectively.

[0025] Furthermore, in this context, providing customized eyeglass frames means providing eyeglass frames that are machined to fit the user's head so that the lens position relative to the user's pupil is adjusted, allowing the lenses to provide the desired and optimal optical correction for that particular user. In other words, a customized eyeglass frame is one that holds the lens in a vertical position relative to the wearer's pupil, providing maximum vision correction.

[0026] Therefore, the present invention provides a method for adjusting at least one frame measurement value in order to ensure maximum vision correction for the wearer. In particular, the frame measurement value to be adjusted is a frame measurement value that complements the control of the lens position with respect to the pupil.

[0027] The lens height represents the distance between the lowest part of the lens and the point on the lens corresponding to the height of the pupil center. To ensure that the lens provides desired vision correction, it is important to guarantee that the point on the lens corresponding to the height of the pupil center is aligned with the actual pupil center.

[0028] Therefore, adjustment of the lens height is important to ensure that the lens provides the necessary optical correction to the user. In particular, in order to minimize the effects of aberration and prism effect while performing maximum vision correction, the vertical position of the optical center of the lens with respect to the pupil needs to be accurate according to the user's prescription. Since each user's face is usually different from another user's face, adjustment of the lens height position is important to ensure that the lens provides the optical correction required for each user.

[0029] The set of facial measurement values preferably includes a plurality of nose measurement values representing the 3D structure of the nose. Since the nose is the main support location for the front of the eyeglass frame, i.e., the frame bridge, the nose is an important facial feature with respect to ensuring that the eyeglass frame is properly seated on the user's face. Therefore, it is important to accurately represent the size and shape of each user's nose so that the frame bridge can be properly adjusted and sized for each user and the eyeglass frame and lens can be correctly positioned with respect to the pupil on the user's face.

[0030] The way the eyeglass frame is positioned on the wearer's nasal bridge affects the lens height and where the lens center is positioned relative to the wearer's pupil center. Determining a set of facial measurements that encompass multiple nasal measurements representing the 3D structure of the nose helps predict how the eyeglass frame sits on the wearer's nasal bridge, and subsequently where the lens center will be positioned relative to the wearer's pupil center. This prediction allows for appropriate adjustment of the lens height to achieve the desired optical correction achieved by the eyeglass frame.

[0031] Preferably, multiple nose measurements encompass multiple rim-to-rim distance (DBR) measurements, with each DBR measurement obtained at a different y-axis point along the length of the nose. Obtaining multiple measurements along the length of the nose helps to construct a more accurate representation of the nose shape. The bridge of the nose is the part of the nose that supports the eyeglass frame, and therefore understanding the structure of the bridge of the nose is important so that the eyeglass frame can be properly adjusted. Thus, multiple nose measurements on the y-axis can help determine a more accurate structure of the bridge of the nose. The y-axis refers to the vertical or longitudinal axis relative to the user.

[0032] There is a correlation between how the eyeglass frame sits on the wearer's face and the position where the lens is held relative to the pupil. If the frame does not sit correctly on the nose, the lens will not be properly held relative to the pupil center. An accurate representation of the nose structure allows for accurate prediction of how the lens will be held in the frame relative to the wearer's pupil, particularly regarding lens height, tilt, and vertex distance. If the prediction indicates that the optical correction tailored to the wearer is suboptimal, appropriate adjustments are made to the frame to adjust the lens height measurement so that the lens is repositioned and aligned with the pupil center at the correct tilt and vertex distance.

[0033] In several examples, multiple nose measurements can be obtained by irradiating the nose along the x-axis at a first y-coordinate (based on Cartesian coordinates) to acquire nose coordinate data and storing it in a nose dataset; adjusting the first y-coordinate by the interval to the second y-coordinate and irradiating the nose along the x-axis at the second y-coordinate to acquire additional nose coordinate data and storing it in the nose dataset; searching for nose coordinate data from the nose dataset; searching for frame bridge data; comparing the nose coordinate data with the frame bridge data; determining the size of the overlapping area between the nose coordinate data and the frame bridge data; and adjusting at least one frame measurement if the size of the overlap is smaller than a threshold. This method allows for a more accurate determination of the seating position of the pivot point of the frame bridge when the eyeglass frame is placed on the user's nose. Sufficient contact area must be provided between the frame bridge and the nose at the seating position so that the eyeglass frame can be worn comfortably without slipping off the user's face. Therefore, this acquisition method provides a good understanding of the structure of the nose and allows the seating position of the frame to be adjusted to each user.

[0034] The comparison step may include mapping a set of frame measurements to a set of facial measurements and identifying a subset of frame measurements that need adjustment. The mapping can be a conceptual map in that data representing the size and structure of the eyeglass frame is compared to data representing the size and structure of the face. In this way, parts of the frame that are not the right size for the user can be identified. Since it is not always the case that all frame measurements need adjustment, identifying a subset of frame measurements that need adjustment means that only the parts of the frame that need correction are adjusted. In this example, the parts of the frame that need adjustment refer to the parts of the frame that adjust the position of the lens relative to the pupil so that the lens provides the optical correction required for this user.

[0035] Preferably, the adjustment step includes adjusting a subset of the identified frame measurements. Thus, it is ensured that only the frame measurements that positively influence the lens position relative to the user's pupil need to be adjusted, rather than adjusting all of them. This positive influence means that the optical center of the lens is positioned well relative to the user's pupil at the appropriate intervertex distance and angle, allowing the lens to provide the best possible optical correction for that user.

[0036] In some cases, the identified subset of frame measurements includes at least one frame measurement related to the bridge of the eyeglass frame. The bridge of the eyeglass frame is the part of the frame that rests on the user's nose. Therefore, the position of the bridge of the eyeglass frame relative to the user's nose determines the position of the lenses relative to their pupils. Thus, by adjusting the frame measurements related to the frame bridge, the position of the frame on the user's nose can be adjusted so as to ensure that the position of each lens relative to each pupil matches the user's prescription.

[0037] The identified subset of frame measurements may include at least one measurement related to the position of various areas of the lens within the eyeglass frame.

[0038] In progressive multifocal lenses (also known as progressive, progressive add-on lenses, or PALs), the power generally changes gradually from the top to the bottom of the lens, providing multiple powers for clear vision at all distances—far, intermediate, near, and somewhere in between. Therefore, it is important to ensure the correct placement of the lens relative to the pupil so that the lens provides the optimal optical correction desired by the particular user.

[0039] Preferably, the step of adjusting the lens position involves separately adjusting the lens position measurement associated with the user's left pupil and the lens height measurement associated with the user's right pupil. In many cases, the user's left and right pupils are not precisely aligned with each other, which means that they are not at the same vertical distance (along the y-axis) on the user's face. Therefore, it is important to measure the positions of the left and right pupils separately so that the positions of the left and right lenses are adjusted accordingly to ensure that each lens is correctly positioned for its respective pupil. Thus, obtaining separate measurements for each pupil ensures that the vertical position of each lens is optimal, meaning that each lens can provide the user with maximum optical correction.

[0040] The step of adjusting at least one frame measurement may include adjusting at least one frame measurement by applying at least one fitting rule to the frame measurement so that the frame measurement satisfies at least one fitting condition. The fitting rule may include comparing the frame measurement to a threshold. Alternatively, or in addition, the fitting rule may include comparing the frame measurement to a facial measurement. Therefore, a fitting rule may be used to determine how much at least one frame measurement needs to be adjusted. Using a fitting rule to supplement the adjustment of frame measurements provides a more accurate method of adjusting eyeglass frames compared to manual adjustment of frame measurements. The fitting rule ensures that the frame is adjusted until the lens is correctly positioned relative to the pupil, reducing the chance of incorrect frame adjustments.

[0041] Preferably, the adjustment step includes applying multiple fitting rules to frame measurements and adjusting the frame measurements so that they satisfy multiple fitting conditions. Using multiple fitting rules to supplement the adjustment of frame measurements further enhances the accuracy when eyeglass frames are adjusted. Different fitting rules may be applied to different frame measurements. This can help ensure that different frame configurations are adjusted to accommodate different frame measurements so that the position of the lens relative to the pupil is correct for the user. One or more fitting rules may be applied to frame measurements. This helps ensure that the frame measurements are correctly adjusted to the user so that the position of the lens relative to the pupil is optimal for the user.

[0042] In some cases, multiple fitting rules may correspond to the nose region, ear region, facial features region, and / or lens region. Multiple fitting rules may encompass some or all of the aforementioned regions. In some cases, multiple fitting rules may encompass alternative and / or additional regions. During use, the eyeglass frame comes into contact with parts of the user's face at several key points where the frame is supported by facial features. These regions include the ears and nose. Therefore, it is important that the eyeglasses are the correct size for each user in these key regions so that the eyeglasses sit correctly on the user's face. This means that the eyeglasses are properly seated when the lens position relative to the user's eye, based on the prescription, provides the maximum vision correction for that user. Thus, establishing several fitting rules related to these key regions helps ensure that a well-fitting eyeglass frame is provided, with the eyeglasses frame being the correct size for each user in these key regions.

[0043] A computer system is provided that adjusts eyeglass frames to a user to provide customized eyeglass frames, comprising: a receiving module configured to receive input data containing three-dimensional coordinate data representing the user's head; a processor configured to identify multiple landmarks in the input data so that the landmarks correspond to the three-dimensional positions of facial features, and to determine a set of facial measurement values ​​based on the multiple landmarks by calculating at least one measurement value associated with at least one landmark using the landmarks; and a search module configured to search a database of frame measurement values ​​for a set of frame measurement values ​​representing a basic eyeglass frame, wherein the processor is further configured to compare the set of facial measurement values ​​with a set of frame measurement values, adjust at least one frame measurement value in the set of frame measurement values ​​based on the comparison, and output a data file containing the adjusted set of frame measurement values, the set of frame measurement values ​​containing lens height measurement values, and the adjustment includes adjusting the lens height measurement value from an initial value to an adjusted value.

[0044] Preferably, the computer system may further include a scan module configured to scan the user's head and generate input data. The scan module may further be configured to transmit the input data to a receiving module. Including the scan module as part of the computer system helps to provide a complete, self-contained computer system for adjusting eyeglass frames.

[0045] In some examples, the scanning system comprises a camera and a signal transmitter. The signal transmitter may be configured to emit multiple signals to the user's head, and the camera may be configured to detect the emitted signals. Thus, a precise method is provided for detecting the user's head and obtaining an image of it, which is then converted into a 3D representation of the user's head in the form of 3D coordinate data. The emission of multiple signals ensures that substantially the entire structure of the user's head is captured.

[0046] Preferably, the signal transmitter is equipped with an infrared transmitter and the camera is equipped with an infrared camera. By capturing infrared images, a convenient and accurate method is provided for acquiring 3D coordinate data of an object.

[0047] In another example, the scanning module is equipped with a LiDAR camera, which provides a convenient and accurate alternative method for obtaining 3D coordinate data of an object.

[0048] According to a third aspect, a computer program is provided that includes instructions causing a computer to perform either the above method or a variation thereof when the computer is executed.

[0049] According to a fourth aspect, a computer-readable data carrier in which the above-mentioned computer program is stored is provided.

[0050] According to the fifth aspect, a computer method is provided for processing image data to estimate facial feature locations, comprising the steps of: receiving input image data that includes three-dimensional coordinate data representing the user's head; preprocessing the input image data to create a preprocessed input dataset; inputting the preprocessed dataset into a first neural network to output a first result, wherein the first result includes three-dimensional coordinate data representing the first estimated location of a facial feature; processing the first result to create a processed input dataset; and inputting the processed input dataset into a second neural network to output a second result, wherein the second result includes three-dimensional coordinate data representing the second estimated location of a facial feature, and the second result differs from the first result.

[0051] Therefore, this method can encompass the sequential use of two neural networks. The first and second neural networks may be the same or different types of neural networks.

[0052] The number of facial landmarks is particularly important in ensuring that eyeglass frames sit correctly on the user's face so that the lenses are properly positioned in front of the pupils. Therefore, accurately identifying the location of these facial landmarks is crucial. The above computer-based approach using machine learning is advantageous because, as a result of improved classification accuracy in the entire algorithm, it is possible to estimate the facial landmark locations more accurately from the input image.

[0053] Accurate interpretation of facial features allows for a more precise representation of facial structure, thereby enabling a more accurate prediction of how the lens will be held within the frame relative to the wearer's pupil. If the prediction indicates that the optical correction is suboptimal for the wearer, the lens height measurement is adjusted, and appropriate adjustments are made to the frame so that the lens is repositioned to align with the pupil center at the correct angle and vertex distance.

[0054] Preprocessing may include sampling input image data to create a first sampled dataset and reducing the dimensionality of the first sampled dataset to create a first reduced dataset. The input image data may further include a color texture map, and sampling may include sampling of the position and color of the input image data. Reducing the dimensionality of the sampled dataset may mean reducing the overall size of the dataset. The first reduced dataset may include three-dimensional coordinate data.

[0055] The process may include sampling the first result to create a second sampled dataset, and reducing the dimensionality of the second sampled dataset to create a second reduced dataset. Sampling may include sampling the position and color of the first result.

[0056] Facial features may include the ear. Facial features may include the pupil.

[0057] A computer-based method for processing image data to estimate the location of facial features (fifth embodiment) can be used in combination with any variation of the first embodiment of the method for adjusting eyeglass frames to fit a user and providing customized eyeglass frames. In particular, the computer-based method for processing image data to estimate the location of facial features can form at least part of the "identification" step of adjusting eyeglass frames to fit a user and providing customized eyeglass frames.

[0058] A computer-based method for processing image data to estimate the location of facial features (fifth embodiment) can be used in combination with a computer system of the second embodiment and any variation thereof configured to adjust eyeglass frames to a user to provide customized eyeglass frames. In particular, when the processor has a configuration for performing the "identification" step, the processor of the computer system may have a configuration for performing the computer-based method for processing image data to estimate the location of facial features.

[0059] According to the sixth aspect, a method is provided for training a classifier for localizing facial features using image data, the method comprising: generating a first training dataset comprising a plurality of head scan images; training a first neural network using the first training dataset, the training comprising: selecting a first subset of the first training dataset and training the first neural network using this subset; determining the accuracy of the first neural network using a second subset of the first training dataset, the second subset of the first training dataset comprising head scan images that do not form part of the first subset; iteratively training the first neural network using the first training dataset; and terminating the training of the first neural network when the difference observed between the outputs of successive iterations is less than a first threshold; and outputting the results of the first neural network, the output representing a first estimate of the location of a facial feature comprising the three-dimensional coordinates and color texture data of the first estimate of the location of a facial feature.The method further comprises generating a second training dataset, wherein the second training dataset includes selected portions of each head scan image from the first training dataset, the selected portions being centered on the three-dimensional coordinates of the first estimated facial feature locations; training a second neural network using the second training dataset, the training comprising the steps of: calculating a set of residuals representing the error of the first neural network; selecting a first subset of the second training dataset and training the second neural network using this subset; determining the accuracy of the second neural network using a second subset of the second training dataset, the second subset of the second training dataset including data that does not form part of the first subset; iteratively training the second neural network using the second training dataset; and terminating the training of the second neural network when the difference observed between the outputs of successive iterations is less than a second threshold; and outputting the results of the second neural network, the output representing the second estimated facial feature locations including the three-dimensional coordinates of the second estimated facial feature locations and color texture data.

[0060] A first training dataset can be generated by observing manually performed facial markings and annotating multiple facial scans with the three-dimensional position coordinates and color texture map values ​​of facial features.

[0061] To avoid concerns, any of the above embodiments or their variations may be combined with one or more of the other embodiments or their variations described. [Brief explanation of the drawing]

[0062] With reference to the attached drawings, embodiments of the present invention are described below merely as examples. [Figure 1a] This shows the xyz coordinate system for a human head based on the Cartesian coordinate system. [Figure 1] This shows the compatible engine API workflow. [Figure 2] An example of a scanning module for computing devices is shown. [Figure 3] Examples of facial features on the front of the face are shown. [Figure 4] An example of a facial landmark on the left side of the face is shown. [Figure 5] Examples of facial measurement values ​​are shown. [Figure 6] Further examples of facial measurements are shown. [Figure 7] Further examples of facial measurements are shown. [Figure 8] An example of frame measurement values ​​is shown. [Figure 9] Further examples of frame measurements are shown below. [Figure 10] Examples of suitability conditions applicable to the nose are shown. [Figure 11] Examples of fitting conditions applicable to the ear are shown. [Figure 12] Examples of suitability conditions applicable to the face are shown. [Figure 13] Further examples of frame measurements are shown below. [Figure 14] Further examples of frame measurements are shown below. [Figure 15] Further examples of frame measurements are shown below. [Figure 16] Further examples of frame measurements are shown below. [Figure 17] This shows a flowchart for advanced DBR calculation. [Figure 18] This shows the structure of the nose divided into layers. [Figure 19] An example of a sampling process using a cylindrical pattern is shown. [Figure 20] An example of a subset of input data is shown. [Modes for carrying out the invention]

[0063] To determine how eyeglass frames will sit on a wearer's nose, knowledge of the shape of the nasal bridge and the frame bridge is necessary to predict the frame's position on the wearer's face.

[0064] This invention relates to a fitting model that can provide an accurate prediction of the frame seating position based on information and data acquired about the frame and the wearer's face, in order to give the ability to determine how the wearer's pupil interacts with the lens when the frame is worn. This prediction can then be used to help provide a well-fitting pair of eyeglasses to the wearer.

[0065] To provide a well-fitting pair of eyeglasses (also called spectacles), standard eyeglass frames are adjusted in size and shape to be unique to the wearer, resulting in a good fit. In particular, certain frame parameters may be adjusted based on the wearer's head measurements to provide a pair of eyeglasses that are properly sized for the wearer's head. We mean that a good fit means that the position of the lenses relative to the wearer's eyes is optimized through the adjustment of the frame that holds the lenses, so that the best vision correction is achieved when the eyeglass frames are placed on the wearer's head. When the eyeglass lenses are properly placed relative to the pupil, the optical axis substantially coincides with the wearer's visual axis. This is important because it means that the prism effect experienced by the wearer when viewing objects through the lenses is minimized.

[0066] To provide a complete set of processed eyeglasses, precise measurements of the subject's facial features, along with accurate measurements of the eyeglass frame elements, must be taken. These measurements can be converted into input data used to mathematically describe the person's head and eyeglass frame. As described in more detail below, this input data can be analyzed and any necessary adjustments made to the frame.

[0067] Eyeglass frames are fabricated using a fitting model that encompasses a set of fitting rules, or a fitting algorithm used to adjust various aspects of eyeglass frames. Different aspects of eyeglass frames correspond to different frame measurements, which will be discussed in more detail below. In particular, the fitting model focuses on four main areas: the nose, ears, facial features, and lenses. Each of these areas is associated with at least one fitting condition, and in some cases multiple fitting conditions, that must be met for the frame to be considered to be the appropriate size for a particular wearer. Therefore, the fitting conditions associated with each area, and algorithms that utilize these fitting conditions through the application of fitting rules, are used to adjust frame measurements and provide customized frames.

[0068] Generally, to adjust the parameters of eyeglass frames, the wearer's head is first scanned using a scanning module to generate a three-dimensional (3D) scan of the wearer's head. The scanned data is output as a 3D data file containing three-dimensional coordinate data representing the wearer's head. Optionally, a depth map and / or texture map may also be output. The 3D data file containing the three-dimensional coordinate data forms the input dataset sent to the processing module. The processing module analyzes the input data of the 3D scan and identifies at least one facial landmark in the 3D scan. The landmark or landmark location corresponds to the three-dimensional position of a facial feature. The landmark location is then used to determine the facial measurement set by calculating the measurements associated with the landmark location. In some cases, the calculation of measurements includes calculating the distance between two landmark locations. In other cases, the calculation of measurements includes calculating the angle formed between two landmark locations and / or the angle formed between the landmark location and a coordinate axis or plane. The facial measurement set generally includes a combination of the above calculations. Therefore, the measurement of facial landmarks based on 3D scan input data can be used to determine the features (e.g., size and shape) of the facial landmarks. The acquisition module retrieves at least one frame measurement from a database of stored frame measurements. The database of stored measurements corresponds to the standardized frame size of a specific part of the frame. The frame measurement is associated with at least one facial landmark. For example, the width of the eyeglass frame is associated with the distance between the wearer's eyes. The acquired frame measurement represents the initial or basic eyeglass frame, which is adjusted and fabricated using a fitting model and applying fitting rules. This step may include comparing the facial measurement to the frame measurement to determine any adjustments to the size or configuration of the basic frame that need to be made to meet the fitting conditions. The fitting conditions are met when the frame measurement is within the threshold limit of the landmark measurement.

[0069] The fitting model rules are used to provide the most suitable frame measurement set proposal to ensure optimal lens positioning relative to the eye when the wearer is wearing the glasses. Frame measurements that meet the fitting criteria in all four main areas of the fitting model are considered optimal for a particular user.

[0070] Further details of the compatible model will be described below. It should be noted that, as shown in Figure 1a, the x-axis refers to the left-right (i.e., horizontal or lateral) direction relative to the user, the y-axis refers to the up-down (i.e., vertical or vertical) direction relative to the user, and the z-axis refers to the front-back (i.e., axial or depth) direction relative to the user.

[0071] Figure 1 shows the overall workflow of the fitting engine API. The fitting engine API can be implemented in a computing system 100, for example, a desktop computer or a mobile computing device. The fitting engine API can also be implemented in a cloud-based system. A user who wants to adjust a frame to fit their head first uploads their eyeglass prescription through interaction with a user interface provided in the computing system (S102). To provide the user with a processed frame, the computing system compares facial measurements with frame measurements and then makes appropriate adjustments to the frame measurements based on the facial measurements and fitting rules.

[0072] Facial measurements are a set of measurements relating to the wearer's face and are obtained solely from the face. Frame measurements are a set of measurements relating to the characteristics of the frame and are obtained solely from the frame. In addition to these sets of facial and frame measurements, an additional set of facial-frame measurements is obtained. These facial-frame measurements are taken on both the frame and the face and are directly related to each other. Finally, a set of lateral measurements is provided, encompassing facial and frame measurements relating to the frame sides and their adjustments. These different sets of measurements form the data to which fitting rules are applied, and a processed eyeglass frame is manufactured. Details of the measurements within each of these sets and how they are obtained are described below.

[0073] A scan module 102, which takes the form of a scanner, is used to acquire facial measurements of the wearer. The scan module 102 is part of a computing system 100, and for example, the scan module 102 may be equipped with a camera 102 of a mobile computing device as shown in Figure 2. Alternatively, the scan module may be another element in communication with the computing system, such as a standalone head scanner. The user scans their head using the scan module 102 to capture data related to the user's head on which facial measurements are being taken (S104).

[0074] To ensure that the entire user's head is scanned, the scan module 102 needs to capture the user's head from multiple angles to ensure that facial data is acquired for the entire user's face. In some cases, the scan module may be positioned to move around the user's head in a fixed position, for example, by drawing a horizontal arc around the user's head from one side of the face, across the front of the user's face, to the opposite side of the face. In other cases, the scan module 102 may remain stationary while the user rotates their head from one side to the other. This latter example is shown in Figure 2. In this example, the scan module 102 is in the form of a mobile device camera, and multiple commands, such as "slowly turn left," instructing the user to move their head from side to side so that a 3D mesh of the user's face consisting of thousands of individual infrared points is created, are presented to the user via the mobile device's display screen. By layering high-resolution scans, an accurate and photorealistic 3D model of the user's face is created. The scan module 102 accurately records the user's head in 3D scans.

[0075] The scan module 102 comprises a camera and a signal transmitter configured to emit multiple signals to the user's face detected by the camera. The emitted signals are used to construct a mathematical image of the user's face.

[0076] In some examples, the scan module 102 takes the form of a TrueDepth® camera equipped with an infrared transmitter. The infrared transmitter projects a number of infrared dots onto the user's face in a known pattern. The infrared camera then captures images of the infrared dots and creates a 3D model of the user's head. In other examples, the scan module 102 takes the form of a LiDAR camera that constructs a 3D model of the user's head using multiple timing laser pulses.

[0077] The 3D scan data is then input to a facial mapping machine learning algorithm that identifies and records certain locations known as landmarks across the head (S106). The machine learning algorithm is pre-trained to identify known facial landmarks from the facial scan data. More specifically, the landmark locations are the locations of prominent facial features. The landmark locations can be recorded as coordinates, for example, xyz coordinates obtained from the 3D scan data. Examples of facial landmarks include the locations of the centers of each pupil, the tip of the nose, and the ears.

[0078] Some facial landmarks are considered relatively more important than others; for example, the position of the ears and pupils are more important than the position of the tip of the nose. This is because, for eyeglass frames to properly sit on the wearer's face so that they hold the lenses in optimal alignment with the wearer's pupils, the arms of the eyeglass frames must be positioned over the ears (in particular, the bent parts of the frame arms must be positioned over the ears), and the lenses must be positioned directly in front of the pupils. Further details regarding the identification of the positions of the ears and pupils will be considered below.

[0079] First, regarding the ear location, the problem with estimating the ear landmark position based on the average ear location measured manually on a head scan set is that the resulting landmark location is generally inaccurate due to the influence of other obstacles such as hair or occlusions, and the difficulty in estimating the position behind the ear in a 180-degree scan.

[0080] To address the technical challenge of accurately and consistently identifying ear landmarks while considering issues with hair and occlusion, an ML algorithm is trained on specific variables using specific data.

[0081] In the following description, we refer to the first and second neural networks. The first neural network, when processed as described below, produces an output that serves as input to the second neural network. The second neural network can use this input to generate highly accurate estimates of the locations of related structures such as the left and / or right ears, and the marker points corresponding to the left and / or right ears. Additionally, certain preprocessing steps are also performed. These steps have been found to improve the classification accuracy of the entire algorithm more than expected, meaning that the algorithm described herein is advantageous because it can estimate marker locations (e.g., ear marker locations) from the input image with greater accuracy.

[0082] First, the 3D scan data (image data) is preprocessed before being input to the first neural network. This preprocessing includes sampling the position and color of the 3D scan data in a cylindrical pattern, as shown in Figure 19. Then, a patch sampling process is performed on the cylindrical sampled position and color data from the 3D scan data, centered on points of initial target landmarks estimated (e.g., manually). The data derived from the patch sampling process has its dimensionality reduced using an orthogonal basis vector set. During training, an orthogonal basis set of 15 vectors is derived from a set of, for example, 3000 patches acquired from a sample head scan (the number of patches can vary, for example, between 1000 and 5000, or between 500 and 10000). The reduction in dimensionality makes it possible to train the system to a useful level of performance with a relatively small number of training examples.

[0083] The input variables to the first neural network are biased according to the mean of the training dataset. Following this biasing step, bias-corrected data is weighted according to the reciprocal square root of the variance observed in the variables of the training dataset. The bias is selected using normalization by the mean and variance of the input data. This bias selection means that the neural network has less information to learn, reducing the network depth and decreasing the total number of iterations required to train the network.

[0084] The preprocessing step helps improve the accuracy of the output for a given amount of training data by assisting the neural network with our prior knowledge about the predicted locations of landmarks (such as ear positions) and by reducing the dimensionality of the data, making the training process more manageable with smaller amounts of annotated head scans.

[0085] The first neural network acts on a low-dimensional representation of a patch of positional and color data that forms part of the input variables. The coefficients of the neural network are adjusted to fit the results of the planned ear markers measured for an existing head scan set. The neural network model used is preferably a multilayer perceptron consisting of a combination of rectified linear units, connected rectified linear units, and Gaussian units. The initial estimates of the ear marker locations are obtained as output in the form of 3D coordinates.

[0086] Once the output from the first neural network is obtained, the patch sampling step is repeated around the position estimate obtained from the first neural network, and the results are input to a second, different neural network that has been specially trained to improve the output from the first neural network. The second neural network outputs the three-dimensional coordinates of landmark points, e.g., left ear and / or right ear landmark points. Since the first neural network has roughly located the coordinates of the landmark points, it will be recognized that the second neural network is acting on a subset of the entire image data that is likely to contain the landmark points. This is advantageous because it improves the final location of the landmarks.

[0087] As described above, using two neural networks sequentially is advantageous because the hyperparameters of the two neural networks can be optimized independently of each other. This can improve the accuracy of the entire marker detection process. In addition, as described above, thanks to the first pass performed by the first neural network, the second neural network acts on a subset of data known to contain ear markers. Thus, it becomes possible to apply domain knowledge that a person's left and right ears are located in roughly obtained locations by the initial neural network, resulting in effective training and implementation.

[0088] We have described the marker detection process using the first and second neural networks, and now we will discuss how to train these neural networks.

[0089] The training data used to train the first neural network is a set of head scans, which will be referred to herein as the first training dataset. Each head scan is essentially an image of the head of a specific person of the type considered above. A suitable number of head scans will be selectable by those skilled in the art relating to this disclosure. Generally, several thousand (e.g., 1000 to 5000, or about 3000) head scans are sufficient. However, the present invention is not limited to using a specific number of head scans for the first training dataset. Each head scan preferably includes geometric data (e.g., in the form of polygonal mesh data), normal vectors at a number of points (or meshes, where relevant) within the geometric data, texture map coordinates, and a color texture map (e.g., an RGB texture map). Each head scan is annotated with three-dimensional left and right ear positions marked using techniques consistent with the observations of trained opticians.

[0090] In a preferred embodiment, a first training dataset is generated by observing ear dot marking as a practice during eyeglass sales and annotating a three-dimensional (e.g., 3000) head scan set at the locations of the left and right ear dots corresponding to this technique.

[0091] The first neural network is trained using a portion of the first training dataset. Once trained, the remainder of the first training dataset is used to measure the accuracy of the first neural network. This process is repeated iteratively until the first neural network is deemed to be producing sufficiently accurate results. In other words, the training phase of the first neural network is terminated when the difference observed between iterations of continuous optimization is less than a threshold. A threshold will be selected by those skilled in the art, relating to the details of the current situation. As an example, the threshold may be set to the root mean square (RMS) of the mean error across six three-dimensional coordinates (i.e., the X, Y, and Z coordinates of the left and right ears) or in that range of 100 microns. However, the threshold may be set to other values ​​instead (e.g., values ​​in the range of 50 to 500 microns), so the present invention is not limited to this value.

[0092] Once the first neural network is trained, the results from the training process are used to generate a second training dataset for training the second neural network. The second neural network is trained using only data that is close to the landmark feature in question (for example, the three-dimensional position coordinates are within a predetermined range of the initial estimates of the landmark location output by the first neural network). In other words, the second neural network is not trained using the entire head scan dataset, but rather using a subset of this data that is particularly relevant to the landmark in question. For example, a subset of data input as shown in Figure 20 (e.g., a 16x16 grid (i.e., 256 samples)), centered on the initial left and right estimates (these estimates are derived from the average locations of the left and right ears obtained across all head scans in the training dataset), along with the three-dimensional coordinates representing the locations of the initial left and right estimates, and a subset of texture data surrounding the initial left and right estimates, forms the second training dataset used to train the second neural network. The second neural network is then trained in the same way as the first neural network, but using the second training dataset.

[0093] During the training of the second neural network, a set of residuals representing the error of the first neural network may be computed. In some cases, residuals can be used to manage the size of the data subset (i.e., the selected portion of the image). For example, if the first neural network outputs an estimate of the ear landmark location at coordinate X+ / -1%, the second neural network can use a subset of the entire image containing the location at coordinate X+ / -1% as its training dataset.

[0094] Alternatively, the second neural network may be trained using the residuals of the first neural network as input parameters. In this case, the second neural network uses the residuals as part of its classification, and the residuals are the input to the second neural network.

[0095] It has been shown that the successive training of such first and second neural networks can result in a particularly accurate marker detection process.

[0096] Now, considering the location of the pupil, the problem with estimating the location of the eye based on the average location measured manually in a head scan set is that, due to the difficulty of the scanning process, this method cannot provide an accurate estimate of the pupil's location. In particular, because the hard surface of the eye cannot be detected in the scan due to the "hole" formed by the pupil, the scan uses the midpoint between the (lateral and medial) outer corners of the eye as the location of the pupil. However, this location is not necessarily equal to the correct pupil location, and there is often variation in the relative pupil placement of each eye.

[0097] To address the technical challenge of accurately and consistently identifying the landmark points (i.e., locations) of both pupils, this invention finds the darkest region of the eye and sets the center of this region as the center of the pupil. This process is as follows:

[0098] As a starting point, the position of each eye is set to the average location identified from manual measurements taken with a head scan set. In this way, an improved approximate location of the pupil is obtained according to the present invention.

[0099] Next, the position and color of the 3D facial scan are sampled in a rectangular pattern centered on the estimated position of each eye. As before, patch sampling is performed to reduce the dimensionality of the sampled position and color data, using the initial eye location estimates as a reference.

[0100] As before, the neural network acts on a low-dimensional representation of the location and color data patches. The coefficients of the neural network are adjusted to fit the results of the planned pupil points measured for an existing head scan set. Similar to ear marker judgment, once an output is obtained, the patch sampling step is repeated using a different neural network specifically trained to improve upon the output from the first neural network. The output of the final neural network is the three-dimensional coordinates of the left and right pupil points.

[0101] It will be recognized that the techniques described above for determining the position of the ears and eyes can also be applied to other facial features, such as the nose, mouth, eyebrows, and lips.

[0102] Figures 3 and 4 show facial features on the front and side of the head. For recognition purposes, the facial features on the right side of the head are the same as those on the left side of the head shown in Figure 4.

[0103] Once landmarks are identified and their corresponding locations are recorded in a landmark coordinate dataset, ray irradiation is applied to the landmark database to extract further information about the facial geometry. This additional information provides details about the size and shape of specific facial features, as well as additional details about how different facial landmarks relate to each other. For example, ray irradiation may be used to determine the shape of the nose based on the shape of the eye, along with the nose landmark data points and interpupillary distance. Determining information about the facial geometry is important because this information is used to construct or construct a virtual image of the individual's head, which is later used to adjust a set of eyeglasses. In this context, the virtual image is a mathematical construct that encompasses a database of data points representing all diverse parts and associated measurements representing the individual's head.

[0104] Since the wearer's head measurements used to adjust frame measurements are determined from this facial data, identifying the location of facial landmarks is crucial. The facial data obtained from the 3D scan must be correctly interpreted so that accurate data relating to the wearer's facial features is entered into the fitting model.

[0105] From a fit perspective, the most important element of facial data is the nose. Since the nose is the facial structure on which the frame rests, it is considered one of the most crucial facial structures for the correct placement and fit of eyeglass frames. The relationship between the frame and the nose largely determines whether the frame fits the wearer's face correctly and provides sufficient visual acuity to improve the wearer's vision through the use of eyeglasses. However, nasal structure varies greatly from person to person, and therefore several variables must be considered to ensure that the eyeglass frame is placed on the bridge of the nose in a way that ensures the lens is correctly positioned relative to the pupil. In particular, the angle of the nose and the bridge of the nose, defined as the ridge formed by the nasal bone itself, vary considerably, and since the nasal bridge is the area on which the eyeglass bridge rests, accurate representation of this area is crucial. The nasal ridge can also take on various shapes, such as straight, concave, convex, or wavy, and these variations should be taken into consideration. Finally, the relative level between the eyes and the bridge of the nose is essential when fitting eyeglass frames, as it affects whether the center of the lens is correctly aligned with the pupil. For the lens position to be properly adjusted, the relationship between the eye and the bridge of the nose must be accurately determined.

[0106] The location of the markers is used to calculate the set of facial measurement values ​​(S108) input into the fitted model, thereby determining the correct position of the frame on the nose and, therefore, the correct position of the lens relative to the pupil. Figures 5, 6, and 7 show some of the various facial measurement values ​​related to nasal markers that can be determined from a 3D head scan.

[0107] Looking at Figure 5 first, the facial measurements related to the shape of the nose are particularly important. In particular, the following facial measurements are judged, and the numbers of the facial measurements correspond to the reference numbers of the measurements in Figure 5. Reference No. 4 in Figure 5—Nasal crest height corresponding to the vertical height in the y-direction at the starting point of the nose. This measurement can be considered to be the starting point of the nose relative to the lower part of the eye (lower eyelid). The upper part or apex of the nasal crest height is known as the nasal crest point. Reference number 5 in Figure 5 – The nasal-frontal angle, which can also be considered the dilation angle, indicates how the nose is expanded or widened when viewed from a forward perspective. This is measured by finding the angle between the outer edge of the nostrils and the central longitudinal axis passing through the nose. Reference number 6 in Figure 6 – The nasal crest angle, which indicates how far the nose protrudes from the face (i.e., how much the nose is pointed upward or downward). This is calculated by finding the angle between the line connecting the nasal crest and the nasal tip and a vertical plane.

[0108] Further details regarding Figure 5 can be found in the appendix.

[0109] Next, looking at Figure 6, prominent facial markers or reference points used to calculate the measurements described with reference to Figure 5 are shown, for example, A(^)CR represents the nasal ridge angle. Further details of Figure 6 can be found in the appendix.

[0110] Finally, Figure 7 shows different nose flaring angles. The flaring angle can also be considered another form of the flaring angle, indicating how the nose expands or widens from an overhead perspective. The flaring angle is calculated by finding the angle formed by the distance between the rims (DBR) of the eyeglass frame at a specified point (which can be considered similar to the frame bridge width), generally 10 mm below the horizontal centerline (the horizontal line formed on the wearer's lower eyelid), and the centerline of the nose. This measurement is important when providing a properly fitted set of eyeglass frames, as the nose-facing surface of the eyeglass frame bridge and the accompanying nose pads should be aligned with the surface of the nose. Therefore, the angle of the frame bridge and the nose flaring angle should be substantially the same for the frame bridge to rest on the bridge of the nose in a way that provides sufficient surface contact area between the nose and the frame, preventing the eyeglasses from slipping down the wearer's nose.

[0111] A basic frame measurement set should also be determined, and an initial dataset of frame measurements should be provided that can be later adjusted to fit the wearer's face.

[0112] A closer examination of the frame measurements suggests that the frame can be divided into several regions, namely the front, side, top, and eye, as seen in Figures 13-16.

[0113] Referring to Figure 13, which illustrates the front of the frame, the frame measurements related to the bridge shape are particularly important. More specifically, the following frame measurements are obtained, and the reference letters below correspond to the reference letters in Figure 13. • d - The inter-lens distance (DBL) corresponding to the distance between the inner edges of the lens apertures. 4—The bridge width line, which corresponds to the horizontal reference line located 5 mm below the horizontal centerline (HCL) (corresponding to the distance on the y-axis when a horizontal line is drawn midway between the tangents to the upper and lower edges of the lens). 5—Bridge width corresponding to the minimum rim-to-rim distance measured along the bridge width line. • 6 – Bridge height measured along the vertical axis of symmetry, corresponding to the distance from the bridge widthline to the lower edge of the bridge. The upper part of the bridge height is defined as the support point.

[0114] Figure 14 shows a side view of a frame from which several frame measurements can be determined, such as the divergence angle AC, which corresponds to the angle formed between the frame bridge and the vertical plane passing through the center of the eyeglass frame. Further details of possible frame measurements are discussed in the appendix.

[0115] Figure 15 shows the top of a frame from which several other frame measurements can be obtained, as discussed in the appendix.

[0116] Figure 16 shows the eye region of the frame from which several further frame measurements can be obtained, such as the forward tilt corresponding to the angle of the lens aperture with respect to the horizontal axis of the eye. Further frame measurements that can be obtained are discussed in the appendix.

[0117] Along with the frame measurements above, additional frame measurements corresponding to equivalent facial measurements are also required. These measurements include the following: • The distance between rims at HCL (0mm); DBR at 5mm; DBR at 10mm; DBR at 15mm – the distance on the x-axis between rims calculated at different positions on the y-axis. Specifically, this is calculated at the nasal ridge point at HCL and at the nasal ridge points 5mm, 10mm, and 15mm above HCL. • Nasal bridge height – the distance on the y-axis between the nasal bridge level (HCL) and the nasal bridge point, which is the midpoint of the nasal bridge. • Spread angle – This also corresponds to the angle between the rear-side plane and the pad plane on the front of the frame. • Frontal angle – Calculated by finding the angle between the vertical line and the plane of the bridge pad, where the vertical line is the central vertical axis at the midpoint of the frame.

[0118] As previously discussed, accurately representing the shape of the nasal bridge and the frame bridge is crucial to determining how the frame bridge sits on the wearer's nose, and subsequently, what adjustments to the frame need to be made to hold the lens in the correct position relative to the wearer's pupil. In other words, accurately predicting the frame's position on the wearer's face makes it possible to determine how the wearer's pupil interacts with the lens when the frame is worn.

[0119] To provide an accurate prediction of the frame seating position, several other important measurements can be determined based on the frame and facial measurements described above.

[0120] A modified form of DBR measurement, performed on the face and known as advanced DBR, allows for a more accurate determination of the seating position for the frame bridge's pivot point when placed on the nose. Referring to Figure 17, these facial measurements are roughly calculated by irradiating the x-axis at predetermined intervals on the y-axis to obtain many measurements along the length of the nose, thus providing a more accurate representation of the nose's size and shape. By performing a good simulation of how the frame will seat on the face, particularly the nose, it is possible to more accurately predict the placement of the spectacle lenses in front of the eyes.

[0121] As shown in Figure 18, the nose can be considered to be formed from multiple nasal layers 180. Each nasal layer is thought to correspond to a vertical slice of the nose, and thus each nasal layer is associated with a specific z-axis coordinate. The layers are generally equally spaced apart along the z-axis so that each nasal layer generally has the same thickness. The number of nasal layers associated with a particular nose may depend on the size of the nose. Each nasal layer is roughly triangular in shape, and the triangle has sides defined by x-axis and y-axis coordinates, and the triangle is located at a certain x-axis coordinate. Each nasal layer forms part of an advanced DBR measurement, and the overall structure of the nose can be constructed from these layers.

[0122] Light irradiation begins at the nasal ridge and ends at the nasal tip. Starting with the y-coordinate corresponding to the nasal ridge (S200), the light is irradiated along the x-axis, and the z-axis positions of different parts of the nose on the x-axis can be determined. The combination of the determined z-axis coordinate data and various x-axis positions for a given y-axis position forms a nasal coordinate dataset that can be stored in the nasal dataset. The y-coordinate is then adjusted at regular intervals (e.g., 1 mm intervals) (S202), and the light irradiation process is repeated to acquire new nasal coordinate data with the new y-coordinate. Additional nasal coordinate data is added to the previously recorded nasal data in the nasal dataset. The light irradiation process is repeated along the length of the nose by adjusting the y-coordinate until it corresponds to the nasal tip and collecting nasal data with the new coordinate (S204).

[0123] Once all nasal coordinate data is obtained, the nasal dataset includes a 3D coordinate dataset representing the shape of the nose (S206). In other words, the advanced DBR measurements include a 3D data cloud of points encompassing multiple nasal layers representing the structure of the nose (e.g., size and shape). Then data corresponding to the shape of the frame bridge is retrieved (S208), also known as the frame bridge data. As previously described with reference to Figure 13, this data includes the spread angle, bridge width line, bridge width, and bridge height.

[0124] The framebridge data includes a 3D coordinate dataset representing the shape of the framebridge. To determine whether the framebridge has the appropriate shape for the wearer's nose, the framebridge is projected onto the nose, and the resulting fit can be evaluated. In particular, both the structure of the nose and the structure of the framebridge are conceptually mapped in 3D, and the 3D maps corresponding to the xyz coordinate dataset represent their respective structures. These structural maps are compared with each other to determine how the framebridge sits on the nose (S210).

[0125] From this comparison, it is possible to determine the amount of overlap between the surface area of ​​the frame bridge and the surface area of ​​the nose, i.e., the size of the contact area, using the nose coordinate data and the frame bridge data (S212). The amount of overlap is a key metric when evaluating whether the eyeglass frame is sized appropriately for a particular wearer.

[0126] On the z-axis, the lens should be as close to the eye as possible, minimizing the distance from the cornea, known as the vertex distance (VD), sometimes called the posterior vertex distance (the wearer's eyelashes should be just above the posterior surface of the lens). The starting point of the lens position should be aligned with the testing equipment used for refraction, and is generally set to a VD of 12.5 mm.

[0127] By projecting the bridge shape of the frame onto advanced DBR measurements, it is possible to detect the contact position, which is the point where the frame meets the nose. To prevent the eyeglass frame from slipping down the nose, a sufficient surface contact area must be provided at the contact position.

[0128] After comparison, if the contact area is too small, i.e., smaller than the threshold specified by the British standard BS EN ISO 12870:2014, the position of the bridge on the nose is adjusted (S214) to increase the contact area, ensuring that the eyeglass frame does not slip off the wearer's nose. In particular, adjustments are made to the DBR, frontal angle, nasal crest angle, and divergence angle at various points on the nose. The final contact position (in xyz coordinates) is then sent to the fitting engine API as the most likely physical position of the frame bridge.

[0129] When determining the most likely physical position of the frame bridge on the nose, it is important to ensure that other parts of the eyeglass frame (e.g., the lower edge of the rim) do not come into contact with parts of the wearer's face (particularly their cheeks and forehead). Since contact between the frame and these parts of the wearer's face is uncomfortable for the wearer, these contact points can be referred to as undesirable contact points.

[0130] To first detect and secondarily avoid these undesirable contact points, the frame bridge is initially positioned at the nasal ridge (where the first high DBR measurement is obtained). The z-axis coordinate of the frame bridge corresponds to the nasal layer coordinates. Since the x-axis position of the frame bridge is the average x-axis position of the nose, the frame bridge is initially placed on the nose.

[0131] To detect undesirable contact points, the frame bridge position is translated along the y-axis until the coordinates representing the frame bridge position pass through the nose layer (corresponding to the coordinates representing the outer surface of the nose, as seen in Figure 18). At this point, undesirable contact occurs. The position of the frame edge may then be adjusted left or right along the x-axis to determine where the frame contacts the wearer's cheek. This process is repeated at each position along the z-axis. A similar technique is used to determine if there are any points on the nose where the top of the spectacle frame contacts the wearer's forehead. By considering these undesirable contact points, the final frame size is determined that ensures the frame does not come into contact with any part of the wearer's face, and if the spectacle frame is prevented from properly seating on the wearer's nose, the lenses will not be optimally aligned with the wearer's pupils. Thus, by taking numerous measurements along the length of the nose, a more detailed assessment of the structure of the nose and frame bridge is made, so that the contact points are determined more accurately and subsequently adjusted to fit the wearer.

[0132] Therefore, advanced DBR measurement is essential when providing well-fitting eyeglass frames that are sized to the wearer's head. In particular, advanced DBR allows for accurate prediction of the frame's position on the wearer's face during use, enabling accurate estimation of the lens's position relative to the pupil during use. Thus, precise adjustment of the frame can be made to adjust the lens's position relative to the pupil so that the lens provides optimal optical correction for the wearer.

[0133] Some nose bridge shapes position the bridge of the nose above the frame's pivot point. To compensate for this effect, a pivot offset is calculated. The pivot offset is the difference between where the frame bridge should touch the nose and where it actually touches the nose. Each frame has a pivot offset that can be calculated so that the correct position of the frame relative to the seating position, which may also be called the contact position, can be determined by the compatible engine API. The seating position is calculated as the sum of the pivot position and the pivot offset.

[0134] Before the eyeglass frame is adjusted, a starting point or basic frame that can be modified is required. Since the frame bridge is an important feature of the frame to be customized, a universal bridge measurement that can be processed later according to each individual is first provided. The universal bridge is a general-purpose bridge shape. To provide a processed frame pair, the universal bridge shape is used as the starting point and adjusted to fit a specific individual based on facial measurements. The adjustment of the bridge improves the seating position of the frame on the wearer's nose by ensuring that the frame bridge is sufficiently supported by the wearer's nose so that the glasses are stable on the wearer's face. In particular, based on the corresponding facial measurements obtained from a facial scan, the spreading angle and the nasal ridge angle of the frame bridge can be adjusted. The pad height is defined by satisfying the conditions related to the nasal ridge height. For example, when the nasal ridge height <x, the pad height = ymm is set, where the height is measured on the y-axis and x represents the average or general value of the nasal ridge height in the population. It is important that the part of the frame that touches the nose has a large contact area, so that the pad height can be adjusted to ensure that there is sufficient surface area between the eyeglass frame and the nose to stabilize the frame on the wearer's face.

[0135] A well-fitted frame must hold the lens in a position that provides the best possible vision correction for a particular wearer. Therefore, it is important to consider the position of the lens during the frame adjustment process to ensure that the lens optical element is optimal for the wearer.

[0136] Adaptation models can be applied to several diverse lens types, including single-focus, progressive, bifocal, and non-prescription lenses.

[0137] A crucial measurement when determining a lens's ability to provide appropriate correction is lens height. In other words, it is essential to provide correct vertical alignment of the lens's optical center relative to each pupil in order to minimize the effects of aberrations and prisms. Therefore, while this is particularly important in the case of high-index lenses and progressive / multifocal lenses—the latter requiring precise measurement of the "wearing" lens position—the lens height position relative to the center of the pupil is generally considered an essential measurement to ensure that the lens provides the optimal correction needed by the wearer and that the relative lens power or segment is aligned with the wearer's pupil.

[0138] As shown in Figure 8, lens height is calculated by measuring the distance on the y-axis between the lowest tangent to the lens edge and the center of the pupil. Therefore, lens height is independent of the overall vertical length of the lens and represents the distance between the lowest part of the lens and a point on the lens corresponding to the height of the center of the pupil. Since the left and right eyes are not necessarily at the same height on a person's face, this measurement is not necessarily the same for both eyes, and therefore separate measurements are obtained for each eye.

[0139] When the pupil corresponds to the optical center of the lens, no forward tilt should be applied. The added lens tilt alters the spherical power of the lens, resulting in a pronounced undesirable cylindrical component. While the power effect is negligible for low-power lenses, it can be quite noticeable for high-power lenses. To avoid negative effects on the millimeter-percentage optical performance of a single-focal lens with its optical center below the wearer's line of sight, the forward tilt should be 2 degrees. For example, as shown in Figure 9a, the optical center is 4 mm below the lens, requiring an 8-degree forward tilt relative to the facial plane.

[0140] As explained, several measurements are taken on both the face and the frame and are directly related to each other. These measurements include BDR, nasal bridge height, bridge height, frontal angle, spread angle, head width, and temple width. Measurements taken only on the face but directly related to one or more frame measurements include vertex curvature radius and nasal bridge angle. Measurements taken only on the frame but directly related to one or more facial measurements include the distance between the pad centers of the frame. Frame measurements related to the frame sides and their adjustments include setback angle, length to flex point, drop length, and total length of the side.

[0141] Once all the above measurements that define the facial measurement set are obtained, and the basic frame measurement set is retrieved from the standard frame measurement database, a process is carried out to adjust the frame measurements to fit the wearer's head perfectly.

[0142] The base frame is selected by comparing the wearer's facial measurements with the corresponding measurements of available base frames and finding the one with the closest initial match between the facial and frame measurements.

[0143] As will be discussed in more detail below, once a basic frame is obtained from the database, fitting rules are applied by a fitting model to adjust the basic frame to suit a specific user and provide a processed eyeglass frame (S110). Once the fitting rules are applied, these can be used to define fitting suggestions for the wearer's frame (S112). A fitting suggestion is a set of frame measurements or size suggestions calculated by applying the fitting conditions (S110). In particular, the suggestions are defined as the style of the frame and the size of these frames, and the position of the lens relative to the pupil satisfies the fitting conditions for four main areas.

[0144] Next, we will examine four main areas that encompass the fitting rules used to present a list of frame options to the wearer.

[0145] optical center position The frame ensures that the optical center of the lens is positioned optimally based on the following: • Z-axis position. Position the lens as close to the eye as possible to maximize the field of view and minimize the VD affected by the eyelashes. • x-axis position. The optical center is positioned directly in front of the pupil at the minimum horizontal center position. • Y-axis position. The vertical center should be at or near the HCL, the height calculation for the progressive lens (generally 22.0 mm) should be minimized, and the optical center should be positioned directly in front of the pupil.

[0146] Then, as shown in Figure 10, the conditions for the frame position to fit the nose can be determined.

[0147] Lens size Lens size affects the overall frame size and is therefore a crucial factor in determining frame size compatibility.

[0148] With regard to facial features, the applicable fitting rule is that good fit of eyeglass frames to facial features is determined by comparing the geometric center distance of the frame with the interpupillary distance of the wearer along its axis—the former should be at least the same size as the former.

[0149] The length-to-length (LTL) width is also compared to the wearer's sphenoid bone width along that axis and serves as a starting point for evaluating fit.

[0150] The requirements for matching the lens and frame width to the face are shown in Figure 12.

[0151] While influenced by the maximum threshold (based on the maximum blank size, weight, and comfort of the lens, or the length-to-depth ratio based on the relative position of the cheekbones), wearers may want to increase the size of the lenses and frames based on aesthetic or stylistic considerations.

[0152] Lens angle • Ensure that the forward tilt at the optical center is zero degrees. • The "wearing position" angle on the side is matched to the wearer's inclination. This requires a more accurate representation of the relative positions of the ears.

[0153] Refer to Figure 9.

[0154] Evaluating the correct temple length Next, considering the ear region, the fitting rule applicable to this region is that good fit to the ear is determined by comparing the frame position with the ear bending point position on the x-axis. Then, as shown in Figure 11, the fitting conditions for the frame position to the ear can be determined. Since the optimal temple length is sufficient, the length to the bending point is at least the same as the length on the x-axis from the back of the lens plane to the starting point of the ear.

[0155] Head fit evaluation • Estimate head width, temple width, and receding angle. • To allow for lateral gripping, the frame HW is estimated as a "corrected" measurement that is approximately 10 mm smaller than the measured HW. The temple width of the frame is measured between the sides 25 mm below the flat front surface. • The setback can be a measurement calculated between the head width and the temple width, after ensuring that the point of closure with the side of the wearer's head is not included.

[0156] Next, by applying each of these rules, the basic frame measurements are adjusted until the fit conditions are met, resulting in a well-fitting frame for the wearer. The fit model then provides a set of frame options that include at least one spectacle frame option based on the optimized basic frame, and the frame options are categorized according to the fit conditions. The wearer can then select a frame from the frame set options (S114).

[0157] The fitting model rules are used to provide a set of frame sizes that are best suited to the wearer, ensuring that the lens position relative to the eye is optimal. The wearer is presented with a list of frames, including those not recommended, via the display screen of the computing system 100, along with the frame suggestions. Frames that meet the fitting criteria (e.g., good and perfect) for all four main areas (nose, ears, facial features, and lenses) are presented to the wearer as frame suggestions. Frames that do not meet the fitting criteria for any of the four main areas are presented to the wearer as not recommended. Frames that meet at least one fitting criterion but fail to meet at least one other fitting criterion are presented to the wearer as average fit. The frames in the frame list are ordered based on how strictly they meet the fitting criteria.

[0158] In summary, a fitted model is used to guide the production of eyeglass frames with specific modifications. The dimensions of various parts of the eyeglass frame are adjusted until all fitting conditions are met, including the nose, ears, facial features, and lens fitting rules.

[0159] Once the wearer has selected their frame, they can choose to further customize the frame from an aesthetic standpoint (S116). Once the wearer has decided on the final frame, the fitting model API accesses the prescription uploaded by the wearer to the computing system (S118) and selects lenses corresponding to the wearer's prescription (S120). The fitting model API then generates a frame specification file (S122) along with a confirmation of the final frame and lenses (S124). The frame specification file is then sent to the eyeglass frame manufacturer (S126), who assembles the wearer's eyeglasses.

[0160] As those skilled in the art will recognize, any and all of the steps described herein can be implemented by a computer having at least a processor and memory. Facial landmarks are output, for example, to a display, transmitted over a network to another device, or printed. In some cases, the output by a 3D printer may be used in combination with other data to print a set of glasses that have been specifically measured and designed for a user's face.

[0161] In some cases, the present invention can also generate color and / or style suggestions using the machine learning techniques described above. Generally, the same process described for identifying landmark locations (particularly ears and eyes) will be used to generate color and / or style suggestions, but the output will be different.

[0162] In short, to create a frame style proposal, three dimensional estimates (e.g., mesh shape) are obtained for one or more landmark locations from a neural network acting on the cylindrical projection depth map representation of the scan data, as previously described. Landmark locations may include (but are not limited to) the eye locations, eyebrow locations (which can be measured at five points on each eyebrow), cheek locations, and mandibular contour locations (which can be measured at five points from the tip of the chin to the bottom of the ear). Shape data is then extracted from coordinate data around the landmark locations (e.g., 5 mm from the periphery of the landmark location) to obtain curvature estimates at each point. The dimensionality of the landmark locations is reduced using example data from a head scan set as a basis. Here, 10 basis vectors are selected to be sent for the ongoing analysis to maximize the variance explained by the basis vectors. Based on measurements of the frame's curve at numerous points around it, the design of the frame to be manufactured is reduced to a set of numbers. The dimensionality of the frame design is reduced using example frame design sets as a basis. The neural network operates on low-dimensional marker location and frame design data. The neural network is tuned to fit sample score sets from a volunteer population. The output of the neural network is a frame style (e.g., in the form of numerical scores) for each individual.

[0163] In short, to create frame color proposals, input data is acquired that includes three-dimensional coordinate data (e.g., in the form of a mesh) and two-dimensional images (e.g., color texture maps) representing the colors at different points on the mesh. Three-dimensional estimates of marker locations are obtained from a neural network acting on the depth map representation of the scan data. Two-dimensional estimates of marker locations in the texture data are obtained by projecting the previously acquired three-dimensional estimates onto the texture image. The texture image is sampled to acquire color data at each marker point. Marker points used for the color proposals include the forehead, left and right cheekbones, eyes (e.g., a 5mm diameter circle centered on the pupil), and hair (obtained as a point 20mm above the ear point). The neural network acts on the extracted color patches and the HSV values ​​of the colors. This neural network is tuned to fit example score sets from a volunteer group. The output of the neural network is the frame color (e.g., in the form of a numerical score) for each individual.

[0164] Note Figures 5 and 6 show several different facial measurements that can be determined from a 3D head scan.

[0165] Regarding Figure 5, the following facial measurement values ​​are determined, and the numbers of the facial measurement values ​​correspond to the reference numbers of the measurements in Figure 5. 4. The apical radius corresponding to the radius of the arc at the top of the nose. 5. The horizontal facial centerline (F-HCL) corresponds to the horizontal line across the nose, which is roughly located on the lower eyelid (lower edge). 6. Facial rim distance (F-DBR) at HCL, which is substantially the same as F-HCL. F-DBR measurements were obtained at different positions on the nose, e.g., 5mm, 10mm, 15mm, and these distances represent the nasal offset (mm) from F-HCL. 7. This is the nasal crest height corresponding to the vertical height in the y-direction, and is the starting point of the nose relative to the F-HCL ("nasal crest"). This measurement can be considered as the point where the nose begins relative to the lower part of the eye (lower eyelid). The upper part or apex of the nasal crest height is known as the nasal crest point. 8. The nasal-frontal angle, which can also be considered the dilation angle, indicates how the nose is expanded or widened from a forward viewpoint. This is measured by finding the angle between the outer edge of the nostrils and the central longitudinal axis of the nose. 9. The nasal crest angle, which indicates how much the nose protrudes from the face (i.e., how much the nose is pointed upward or downward). This is calculated by finding the angle between the line connecting F-HCL and the nasal tip and a vertical plane. 10. The length of the facial flexure that substantially corresponds to the distance between the bridge of the nose and the ear. In other words, this is the distance of the portion of the spectacle frame that starts from the hinge of the frame and ends at the starting point of the flexure of the frame arm. 11. Facial head width, generally measured as the distance between the wearer's ears and at a height corresponding to the top of the ears, corresponds to the width of the wearer's head on the x-axis. Head width is measured between "ear points" and corresponds to the distance between the midpoints of each bend in the frame, not the point of maximum head width. Frame head width is usually "corrected" to be approximately 10 mm smaller than the measured head width so that side grips can be provided. 12. Facial temple width corresponding to the x-axis distance between the wearer's left and right temples at the side of the head. This measurement is related to the sphenoid bone width, which is the x-axis distance between the left and right sphenoid bones (roughly at the edges of the eyebrows). The optical standard specifies the temple width of the frame measured between the sides 25 mm below the front plane. 13. Interpupillary distance, which corresponds to the distance between the center points of each pupil.

[0166] Next, looking at Figure 6, we see prominent facial markers or reference points used to calculate the measurements described with reference to Figure 5. In particular, the following reference points are marked, and the letters of the reference points correspond to the reference letters of the features in Figure 6. • S - Sphenoid bone related to the width of the facial temples and the width of the sphenoid bone • T - Temples related to facial temple width A - Ear flexion points related to facial flexion length and facial head width I - Length of the bend • LP – Pupillary centerline related to interpupillary distance • Facial horizontal centerline (H-HCL) related to LC-F-HCL and F-DBR ·A(^)CR―Nasal ridge angle ·i - degree of anterior inclination • PM - Frame Plane

[0167] Referring to Figure 13, which shows the front of the frame, the following frame measurements were performed, and the reference characters below correspond to the reference characters in Figure 13. • CR, CL – Right and left centers corresponding to the center points of the right and left lens openings of the frame. • a - Horizontal lens size corresponding to the width of the lens aperture • b - Vertical lens size corresponding to the height of the lens aperture • d - Inter-lens distance (DBL) corresponding to the distance between the inner edges of the lens apertures. • 1—The horizontal center line (HCL) of the frame, corresponding to the distance on the y-axis from the midpoint of the horizontal line between the tangents of the upper and lower edges of the lens. • 2R, 2L – Right and left vertical center lines corresponding to the vertical line passing through the center point of each lens aperture. 3. Vertical axis of symmetry on the front of the frame • 4—Bridge width line corresponding to the horizontal reference line located 5mm below the HCL • 5 - Bridge width corresponding to the minimum distance between rims measured on the bridge width line • 6 – Bridge height measured along the vertical axis of symmetry, corresponding to the distance from the bridge widthline to the lower edge of the bridge. The upper part of the bridge height is defined as the support point. • 7 - Lens width (LTL) calculated as (lens width × 2 + bridge width - 1)

[0168] Referring to Figure 14, which illustrates the side of the frame, the following frame measurements can be obtained, with the reference letters below corresponding to the reference letters in Figure 14. 1—This is the axis of the hinge joint (e.g., a dwell screw), and the joint connects the frame arm to the front of the frame. 2. Midline of the hinge joint • 3 - Side centerline corresponding to the horizontal line passing through the central vertical axis of the frame arm. 4. The length of the bend corresponding to the distance between the hinge joint and the bend of the frame arm. 5 - The downward length corresponding to the length of the frame arm portion extending from the bent part of the frame arm. • I - The total side length corresponding to the sum of dimensions 4 and 5 (i.e., the sum of the bend length and the descent length) • Details of the measurement location of the X-joint • FTB – Front bend length corresponding to the distance between the rear surface of the front of the frame and the side bend. • AC – Spread angle corresponding to the angle formed between the frame bridge and the vertical plane passing through the center of the eyeglass frame.

[0169] Referring to Figure 15, which shows the top of the frame, the following frame measurements are obtained, with the reference letters below corresponding to the reference letters in Figure 15. ·λ—The receding angle corresponding to the angle between the side (frame arm) and the normal to the front of the frame.

[0170] Referring to Figure 16, which shows the eye region of the frame, the following frame measurements can be obtained. The intervertex distance (VD), sometimes called the posterior vertex distance (BVD), corresponds to the distance between the point on the frame representing the lens's viewpoint (the point where the wearer's eye's visual axis intersects the back surface of the corrective eyeglass lens) and the pupil. The intervertex distance is measured along the visual axis. • Forward tilt corresponding to the angle of the lens aperture relative to the horizontal axis of the eye • Forward tilt angle corresponding to the angle of the front of the frame relative to the temples [Explanation of Symbols]

[0171] 1. Frame horizontal centerline (HCL) / Hinge joint axis 2 Hinge joint midline 2R,2L Right vertical center line, left vertical center line 3. Vertical axis of symmetry / lateral center line 4. Nasal bridge height / apical radius / bridge width line / bend length 5. Nasal frontal angle / Facial horizontal centerline (F-HCL) / Bridge width / Descending length 6. Nasal ridge angle / distance between facial rims (F-DBR) / bridge height 7. Nasal bridge height / lens width (LTL) 8 Nasofrontal angle 9 Nasal ridge angle 10. Length of the facial flexion 11. Face-to-head width 12 Face Temple Width 13 Interpupillary distance S Sphenoid T Temple A. Ear flexion point I. Length of the bent section / Total length of the side L,P. Pupil center line LC (F-HCL) - Horizontal centerline of the face ACR nasal ridge angle i Anterior slope PM frame plane CR,CL Center right, Center left a Horizontal lens size b Vertical lens size d Interlens distance (DBL) Details of the measurement location of the X joint FTB front bend length AC spread angle λ Receding angle

Claims

1. A computer-based method for providing customized eyeglass frames by adjusting them to suit the user, The process involves receiving input data that includes three-dimensional coordinate data representing the user's head, The step of identifying multiple landmark locations within the input data, wherein the landmark locations correspond to three-dimensional locations of facial features, A step of determining a set of facial measurement values ​​based on the plurality of landmark locations by calculating at least one measurement value associated with at least one landmark location, The steps include: searching a database of frame measurement values ​​for a set of frame measurement values ​​that represents a basic eyeglass frame; The steps include comparing the facial measurement set with the frame measurement set, A step of adjusting at least one frame measurement value in the frame measurement value set based on the comparison, A step of outputting a data file containing the adjusted frame measurement set, It includes, The frame measurement set includes lens height measurements, and the adjustment step includes adjusting the lens height measurements from an initial value to an adjustment value. The aforementioned set of facial measurement values ​​includes multiple nasal measurement values ​​that represent the 3D structure of the nose. The aforementioned multiple nasal measurements encompass multiple inter-rim distance (DBR) measurements, Each DBR measurement was acquired at various y-axis points along the length of the nose. A method for defining lens height as the distance between the lowest point of the lens and a point on the lens corresponding to the height of the pupil center.

2. The process involves irradiating a light ray along the x-axis of the first y-coordinate to acquire nasal coordinate data, and storing the nasal coordinate data in a nasal data set, wherein the x-axis is in the left-right direction relative to the user. The following steps are taken to obtain the multiple nose measurements: adjusting the first y coordinate based on the interval up to the second y coordinate, irradiating the second y coordinate along the x axis to acquire additional nose coordinate data, storing the additional nose coordinate data in the nose dataset, and adjusting the seating position of the eyeglass frame. Here, adjusting the seating position of the eyeglass frame is performed by: The nasal coordinate data is retrieved from the nasal dataset, Searching for frame bridge data, The nasal coordinate data is compared with the frame bridge data, To determine the size of the overlapping area between the aforementioned nasal coordinate data and the aforementioned frame bridge data, The method according to claim 1, further comprising adjusting at least one frame measurement when the size of the overlap is less than a threshold.

3. The method according to claim 1 or 2, wherein the comparison step comprises mapping the frame measurement set to the facial measurement set and identifying a subset of frame measurement values ​​that require adjustment.

4. The method according to claim 3, wherein the adjustment step comprises adjusting the frame measurement subset.

5. The method according to claim 3 or 4, wherein the frame measurement subset includes at least one frame measurement relating to the bridge of the eyeglass frame.

6. The method according to any one of claims 3 to 5, wherein the frame measurement subset includes at least one measurement relating to the position of the lens section within the eyeglass frame.

7. The method according to any one of claims 1 to 6, wherein adjusting the lens height measurement includes separately adjusting the lens height measurement associated with the user's left pupil and the lens height measurement associated with the user's right pupil.

8. The method according to any one of claims 1 to 7, wherein the step of adjusting the at least one frame measurement comprises applying at least one conformance rule to the frame measurement and adjusting the at least one frame measurement such that the frame measurement satisfies at least one conformance condition.

9. The method according to claim 8, wherein the adjusting step comprises applying a plurality of conformance rules to the frame measurement and adjusting the frame measurement so that the frame measurement satisfies a plurality of conformance conditions.

10. The method according to claim 9, wherein the aforementioned multiple fitting rules correspond to the nasal region, the ear region, the facial contouring region, and the lens region.

11. A computer system configured to adjust eyeglass frames to suit the user and provide customized eyeglass frames, A receiving module configured to receive input data that includes three-dimensional coordinate data representing the user's head, It is a processor, Multiple landmark locations are identified within the aforementioned input data, and these landmark locations correspond to the three-dimensional positions of facial features. A set of facial measurement values ​​is determined based on the plurality of landmark locations by calculating at least one measurement value associated with at least one landmark location. A processor configured as follows, A search module configured to retrieve a set of frame measurement values ​​representing basic eyeglass frames from a database of frame measurement values, It is equipped with, The aforementioned processor further, The facial measurement set is compared with the frame measurement set. Based on the above comparison, adjust at least one frame measurement value in the frame measurement value set, Outputs a data file containing the adjusted frame measurement set. It is configured in such a way, The frame measurement set includes lens height measurements, and the adjustment includes adjusting the lens height measurements from an initial value to an adjusted value. The aforementioned set of facial measurement values ​​includes multiple nasal measurement values ​​that represent the 3D structure of the nose. The aforementioned multiple nasal measurements encompass multiple inter-rim distance (DBR) measurements, Each DBR measurement was acquired at various y-axis points along the length of the nose. The lens height represents the distance between the lowest point of the lens and a point on the lens corresponding to the height of the pupil center. Computer system.

12. The computer system according to claim 11, further comprising a scan module configured to scan a user's head and generate input data, and to transmit the input data to the receiving module.

13. The computer system according to claim 12, wherein the scan module comprises a camera and a signal transmitter, the signal transmitter is configured to transmit a plurality of signals to the user's head, and the camera is configured to detect the transmitted signals.

14. The computer system according to claim 13, wherein the signal transmitter comprises an infrared transmitter, and the camera comprises an infrared camera.

15. The computer system according to claim 12, wherein the scan module comprises a LiDAR camera.

16. A computer program that, when executed by a computer, includes instructions causing the computer to perform the method according to any one of claims 1 to 10.

17. A computer-readable data carrier storing the computer program described in claim 16.

Images