3D human ear model construction method, system, device and medium

Abstract

The present application discloses a method, system, device and medium for constructing a 3D human ear model, wherein the method includes: obtaining original point cloud data of a human ear; obtaining voxel features of a human ear according to the original point cloud data and a preset first feature extraction algorithm; obtaining point cloud features of a human ear according to the original point cloud data and a preset second feature extraction algorithm; obtaining bird's-eye view features according to voxel features of a human ear and a preset mapping algorithm; integrating the voxel features of a human ear, the bird's-eye view features and the point cloud features of a human ear with a preset integration algorithm to obtain a feature vector; and constructing a three-dimensional model of a human ear according to the feature vector. The method for constructing a 3D human ear model of the present application combines different point cloud processing methods to extract features, so that the final three-dimensional model of a human ear pays more attention to the tiny and complex structure of the human ear, is more compatible with the human ear, and improves the accuracy of constructing a human ear model.

Description

Technical Field

[0001] The present application relates to the field of human ear detection, and in particular to a 3D human ear model construction method, system, device and medium. Background Art

[0002] With the popularity of the TWS headset industry, consumers have higher and higher requirements for the comfort of wearing headsets. Using human factors engineering to improve the comfort of headsets has become one of the competitive advantages of headset manufacturers. The human ear is a feasible biological feature with characteristics that the human face, iris, etc. do not have. Experiments have shown that the shape of the human ear is unique. Therefore, analyzing the human ear, constructing a corresponding three-dimensional model, and then making headphones through the three-dimensional model has become one of the current trends.

[0003] In the related technology, the relevant algorithms of 3D target detection are applied to human ear detection, but the existing detection algorithms often only focus on the global characteristics of the human ear, resulting in a large gap between the final constructed three-dimensional model and the actual human ear, and the produced headphones do not bring high comfort. Summary of the invention

[0004] The present application aims to solve at least one of the technical problems existing in the prior art. To this end, the present application proposes a method for constructing a 3D human ear model, which combines different point cloud processing methods to extract features, so that the final three-dimensional human ear model pays more attention to the tiny and complex structure of the human ear, is more compatible with the human ear, and improves the accuracy of the human ear model construction.

[0005] This application also proposes a 3D human ear model construction system.

[0006] The application also provides an electronic device.

[0007] The present application also proposes a computer-readable storage medium.

[0008] A 3D human ear model construction method according to an embodiment of the first aspect of the present application includes:

[0009] Get the original point cloud data of the human ear;

[0010] Obtaining a human ear voxel feature according to the original point cloud data and a preset first feature extraction algorithm;

[0011] Obtaining a human ear point cloud feature according to the original point cloud data and a preset second feature extraction algorithm;

[0012] Obtaining a bird's-eye view feature according to the human ear voxel feature and a preset mapping algorithm;

[0013] Integrate the human ear voxel feature, the bird's-eye view feature, and the human ear point cloud feature using a preset integration algorithm to obtain a feature vector;

[0014] A three-dimensional model of the human ear is constructed according to the feature vector.

[0015] The 3D human ear model construction method according to the embodiment of the present application has at least the following beneficial effects: different point cloud processing methods are combined to extract features, so that the final three-dimensional model of the human ear pays more attention to the tiny and complex structure of the human ear, is more compatible with the human ear, and improves the accuracy of human ear model construction.

[0016] According to some embodiments of the present application, obtaining a human ear voxel feature according to the original point cloud data and a preset first feature extraction algorithm includes:

[0017] Dividing the original point cloud data into voxels of equal proportions in space to obtain a plurality of human ear voxels;

[0018] Performing voxel feature encoding on the plurality of human ear voxels to obtain a plurality of voxel-wise features;

[0019] Perform 3D sparse convolution on the multiple voxel-wise features to obtain human ear voxel features.

[0020] According to some embodiments of the present application, the human ear voxel includes: a first voxel and a second voxel, the first voxel being characterized as a screened human ear voxel, and the second voxel being characterized as an unscreened human ear voxel;

[0021] The 3D human ear model construction method further includes:

[0022] The second voxel is screened according to a preset voxel interval to obtain the first voxel.

[0023] According to some embodiments of the present application, the preset second feature extraction algorithm includes: a preset scale space extreme value retrieval algorithm, a preset T-net algorithm, a preset MLP network algorithm and a preset aggregation algorithm;

[0024] The step of obtaining the human ear point cloud feature according to the original point cloud data and a preset second feature extraction algorithm includes:

[0025] The original point cloud data is searched using the preset scale space extreme value search algorithm to obtain a spatial extreme value range interval;

[0026] Adjusting the original point cloud data according to the spatial extreme value range, the preset T-net algorithm and the preset aggregation algorithm to obtain an adjustment vector;

[0027] Feature extraction is performed according to the preset MLP network algorithm and the adjustment vector to obtain human ear point cloud features.

[0028] According to some embodiments of the present application, adjusting the original point cloud data according to the spatial extreme value range interval, the preset T-net algorithm and the preset aggregation algorithm to obtain an adjustment vector includes:

[0029] Adjusting the original point cloud data using the preset T-net algorithm according to the spatial extreme value range to obtain adjusted point cloud data;

[0030] The adjusted point cloud data is aggregated according to the preset aggregation algorithm to obtain aggregated point cloud data.

[0031] According to some embodiments of the present application, the preset mapping algorithm is a downsampling algorithm.

[0032] According to some embodiments of the present application, the human ear voxel feature, the bird's-eye view feature, and the human ear point cloud feature are integrated using a preset integration algorithm to obtain a feature vector, including:

[0033] The human ear voxel feature, the bird's-eye view feature, and the human ear point cloud feature are sampled respectively to obtain a plurality of key point features;

[0034] The multiple key point features are converted into feature space through a two-layer FCN network to obtain the feature vector

[0035] A 3D human ear model construction system according to the second aspect of the present application includes:

[0036] An acquisition module, wherein the acquisition module is used to acquire original point cloud data of a human ear;

[0037] An extraction module, the extraction module is used to obtain a human ear voxel feature according to the original point cloud data and a preset first feature extraction algorithm;

[0038] The extraction module is also used to obtain the human ear point cloud features according to the original point cloud data and a preset second feature extraction algorithm;

[0039] A calculation module, the calculation module is used to obtain a bird's-eye view feature according to the human ear voxel feature and a preset mapping algorithm;

[0040] The calculation module is further used to integrate the human ear voxel feature, the bird's-eye view feature, and the human ear point cloud feature using a preset integration algorithm to obtain a feature vector;

[0041] A construction module is used to construct a three-dimensional model of a human ear according to the feature vector.

[0042] The 3D human ear model construction system according to the embodiment of the present application has at least the following beneficial effects: different point cloud processing methods are combined to extract features, so that the final three-dimensional model of the human ear pays more attention to the tiny and complex structure of the human ear, is more compatible with the human ear, and improves the accuracy of human ear model construction.

[0043] An electronic device according to an embodiment of the third aspect of the present application includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the 3D human ear model construction method as described in any one of the embodiments of the first aspect is implemented.

[0044] According to the computer-readable storage medium of the fourth aspect embodiment of the present application, the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the 3D human ear model construction method as described in any one of the first aspect embodiments.

[0045] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The present application is further described below with reference to the accompanying drawings and embodiments, wherein:

[0047] Figure 1 A flowchart of a method for constructing a 3D human ear model provided in an embodiment of the present application;

[0048] Figure 2 A flowchart of a method for constructing a 3D human ear model provided in an embodiment of the present application;

[0049] Figure 3 A flowchart of a method for constructing a 3D human ear model provided in an embodiment of the present application;

[0050] Figure 4 A flowchart of a method for constructing a 3D human ear model provided in an embodiment of the present application;

[0051] Figure 5 A flowchart of a method for constructing a 3D human ear model provided in an embodiment of the present application;

[0052] Figure 6 A module block diagram of a 3D human ear model construction system provided in an embodiment of the present application.

[0053] Figure numerals: 100, acquisition module; 200, extraction module; 300, calculation module; 400, construction module. DETAILED DESCRIPTION

[0054] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be understood as limiting the present application.

[0055] In the description of the present application, it should be understood that descriptions involving orientation, such as up, down, front, back, left, right, etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0056] In the description of this application, "several" means more than one, "more" means more than two, "greater than", "less than", "exceed", etc. are understood to exclude the number itself, and "above", "below", "within", etc. are understood to include the number itself. If there is a description of "first" or "second", it is only used for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0057] In the description of this application, unless otherwise clearly defined, terms such as setting, installing, connecting, etc. should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in this application based on the specific content of the technical solution.

[0058] In the description of the present application, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0059] Reference Figure 1 In a first aspect, the present application proposes a method for constructing a 3D human ear model, including but not limited to step S100, step S200, step S300, step S400, step S500 and step S600.

[0060] Step S100: obtaining original point cloud data of the human ear;

[0061] Step S200: obtaining human ear voxel features according to the original point cloud data and a preset first feature extraction algorithm;

[0062] Step S300: obtaining a human ear point cloud feature according to the original point cloud data and a preset second feature extraction algorithm;

[0063] Step S400: obtaining a bird's-eye view feature according to the human ear voxel feature and a preset mapping algorithm;

[0064] Step S500: integrating the human ear voxel features, the bird's-eye view features, and the human ear point cloud features using a preset integration algorithm to obtain a feature vector;

[0065] Step S600: constructing a three-dimensional model of the human ear according to the feature vector.

[0066] The earprint of the human ear is extracted by three-dimensional scanning to obtain the corresponding original point cloud data of the human ear. The original point cloud data is divided into two branches for processing. One branch extracts the features of the original point cloud data by a preset first feature extraction algorithm to obtain the voxel features of the human ear, and the other branch extracts the features of the original point cloud data by a preset second feature extraction algorithm to obtain the point cloud features of the human ear. The voxel features of the human ear are then mapped to a 2D feature bird's-eye view by a preset mapping algorithm to obtain the bird's-eye view features of the two-dimensional bird's-eye view. The voxel features of the human ear, the bird's-eye view features and the point cloud features of the human ear are then integrated by a preset integration algorithm to obtain a feature vector. A three-dimensional model of the human ear corresponding to the human ear is constructed according to the integrated feature vector. After the three-dimensional model of the human ear is constructed, the corresponding headphones can be designed based on the three-dimensional model to improve the comfort of the headphones.

[0067] For example, for universal headphones, different feature vectors are obtained based on the point cloud data of different ears, and then these feature vectors are analyzed and processed (such as cluster analysis by Euclidean clustering), and the size model with a large number of results is used as the reference size of universal headphones. For customized headphones, the three-dimensional model constructed by the obtained feature vectors can provide a reference for modification to make the headphones more compatible with the ears.

[0068] It should be noted that the preset first feature extraction algorithm mentioned in the present application is a voxel-based feature extraction, and the preset second feature extraction algorithm is a point-based feature extraction.

[0069] The 3D human ear model construction method of the embodiment of the present application combines different point cloud processing methods to extract features, so that the final three-dimensional human ear model pays more attention to the tiny and complex structure of the human ear, improves the accuracy of human ear model construction, and is more compatible with the human ear. The headphones made based on the three-dimensional model are also more compatible with the human ear, improving the comfort of the headphones.

[0070] Reference Figure 2 In some embodiments of the present application, step S200 includes but is not limited to step S210, step S220 and step S230.

[0071] Step S210: dividing the original point cloud data into voxels of equal proportion in space to obtain a plurality of human ear voxels;

[0072] Step S220: performing voxel feature encoding on a plurality of human ear voxels to obtain a plurality of voxel-wise features;

[0073] Step S230: performing 3D sparse convolution on multiple voxel-wise features to obtain human ear voxel features.

[0074] Due to the irregularity of the original point cloud data, if the original point cloud data is simply divided equally, the number of point clouds in each voxel will be unequal, and some will even be empty. In order to avoid this situation, the original point cloud data is divided into small cubes (voxels) of equal proportion in space to obtain multiple human ear voxels, each with the same number of point clouds. The obtained human ear voxels are then subjected to VFE (Voxel Feature Encoding) to obtain voxel-wise features. The multiple voxel-wise features obtained are subjected to 3D sparse convolution with the 3D sparse convolution layer in 3D CNN (Convolutional Neural Networks) to obtain human ear voxel features.

[0075] The VFE algorithm can retain the characteristics of the point cloud itself (global features), and can also combine the local structural features of each point cloud within its voxel, so that the final features not only focus on the overall global features, but also focus on the tiny and complex features of the ear, thereby improving the accuracy of model construction.

[0076] When performing 3D sparse convolution on voxel-wise features, since the point cloud data in each voxel is the same, the parameters and weights can be shared during the convolution operation, which can save computational effort.

[0077] In some embodiments of the present application, the human ear voxel includes: a first voxel and a second voxel, the first voxel is characterized as a screened human ear voxel, and the second voxel is characterized as an unscreened human ear voxel.

[0078] The 3D human ear model construction method further includes but is not limited to step S700.

[0079] Step S700: screening the second voxel according to a preset voxel interval to obtain the first voxel.

[0080] After the original point cloud data is divided into equal proportions in space, the point cloud data in each voxel is the same. By using random sampling, the voxels outside the preset voxel interval are discarded, leaving only the voxels within the preset voxel interval.

[0081] For cubes with too small length, too small volume, and the same number of point clouds, it means that the local point clouds are dense, and some data can be appropriately ignored, which does not affect the feature extraction in this area. For cubes with too long length, too large volume, and the same number of point clouds, it means that the features of the model in this area are not representative and can be ignored. By setting this up, the imbalance between voxels is reduced, the sampling bias is reduced, and more variability is added during training.

[0082] It should be noted that the setting of the preset voxel interval needs to be analyzed specifically according to the actual situation. For example, T represents the minimum value of the preset voxel interval, and M represents the maximum value of the preset voxel interval. For in-ear headphones, it is necessary to focus on the feature points near the concha cavity and the ear canal opening. When setting M, the size and depth of the ear canal opening need to be considered for the feature points in this area, so when setting T, it should be set a little smaller. For semi-in-ear headphones, you only need to pay attention to the size of the concha cavity. When setting T and M, the value of M should be slightly larger than the M value of the in-ear type. Or for headphones, what needs to be considered is the size and angle of the auricle. When setting T and M, the set size should be larger than that of in-ear or semi-in-ear headphones.

[0083] Reference Figure 3 In some embodiments of the present application, the preset second feature extraction algorithm includes: a preset scale space extreme value retrieval algorithm, a preset T-net algorithm, a preset MLP network algorithm and a preset aggregation algorithm.

[0084] Step S300 includes but is not limited to step S310 , step S320 , and step S330 .

[0085] Step S310: searching the original point cloud data using a preset scale space extreme value search algorithm to obtain a spatial extreme value range interval;

[0086] Step S320: adjusting the original point cloud data according to the spatial extreme value range, the preset T-net algorithm and the preset aggregation algorithm to obtain an adjustment vector;

[0087] Step S330: performing feature extraction according to a preset MLP network algorithm and an adjustment vector to obtain a human ear point cloud feature.

[0088] The comfort of wearing headphones is mainly related to the shape and size of the concha cavity of the human ear. The design of in-ear headphones must be slightly smaller than the size of the concha cavity. If it is too large, it will compress the ear when worn, causing ear pain, and if it is too small, it will easily cause the earphone to slip off. Therefore, when extracting feature points of the ear, it is necessary to analyze the extreme points of the concha cavity.

[0089] The original point cloud data is retrieved by a preset scale space extreme value retrieval algorithm to obtain a spatial extreme value range interval. In this embodiment, a Gaussian differential function is used to identify potential scale and rotation invariant interest points, and the range of extreme value points is roughly determined to obtain a spatial extreme value range interval. The point cloud data retrieved within the spatial extreme value range interval is adjusted by a preset T-net algorithm to complete the alignment of the feature spaces of different input points, and then the adjusted point cloud data is aggregated by a preset aggregation algorithm to obtain an adjustment vector; the adjustment vector is passed through MLP (Multi-Layer Perception) to obtain a human ear point cloud feature, and multiple human ear point cloud features are combined to obtain a point-based human ear point cloud feature vector.

[0090] By first using a preset scale space extreme value retrieval algorithm to extract extreme point features, a large amount of unnecessary point cloud data is reduced and the amount of calculation is reduced.

[0091] It should be noted that T-net is the network structure in PointNet. PointNet is a point cloud classification / segmentation deep learning framework proposed by Stanford University in 2016.

[0092] Reference Figure 4 In some embodiments of the present application, step S320 includes but is not limited to step S321, step S322 and step S323.

[0093] Step S321: adjusting the original point cloud data using a preset T-net algorithm according to the spatial extreme value range interval to obtain adjusted point cloud data;

[0094] Step S322: Aggregate each point cloud data in the adjustment point cloud data according to a preset aggregation algorithm to obtain an adjustment vector.

[0095] The point cloud data is input into the T-net network, where the T-net network can perform rotational invariance processing on the model, complete the spatial feature alignment of the point cloud data, and then implement aggregation processing of the aligned point cloud data through a preset aggregation algorithm, where the preset aggregation algorithm is a logarithmic function, thereby outputting an adjustment vector that is completely independent of the input order, solving the problem of point cloud disorder and enabling spatial alignment of different human ear data.

[0096] In some embodiments of the present application, the preset mapping algorithm is a downsampling algorithm.

[0097] By downsampling, 3D features are converted into 2D bird's-eye view features to obtain bird's-eye view features. The point cloud can also be divided into several pieces according to height, and the maximum height is used as the feature of the point. In this way, images of several channels can be obtained to obtain bird's-eye view features.

[0098] Reference Figure 5 In some embodiments of the present application, step S600 includes but is not limited to step S610 and step S620.

[0099] Step S610: sampling the human ear voxel features, the bird's-eye view features, and the human ear point cloud features respectively to obtain a plurality of key point features;

[0100] Step S620: convert multiple key point features into feature space through a two-layer FCN network to obtain a feature vector.

[0101] Perform RoI-grid Pooling via Set Abstraction on each feature (ROI grid pooling is obtained by set abstraction, ROI stands for Region of Interest, and each region of interest corresponds to a bounding box) to obtain multiple 3D ROIs, and then propose ROI pooling (Region Of Interesting pooling, ROIpooling, region of interest pooling) for each 3D ROI, sample each region of interest, sample 6x6x6 gridpoints in each region, and obtain G = {g1, g2, ... g216}. Draw a sphere with a radius of R at the center of each grid point to obtain a neighborhood sphere, and then find the key point features based on the neighborhood sphere. Multiple key point features get a set of vectors. By setting multiple different radii R, multi-scale features can be obtained, thereby obtaining multiple sets of vectors. Multiple sets of vectors are input to PointNetBlock (SA layer) and abstract features are obtained through Set Abstraction. Multiple abstract features are integrated through the PoinNet module, and then converted into 256 dimensions through a two-layer fully connected network (Fully Convolutional Networks, FNC) to obtain the final feature vector.

[0102] Reference Figure 6 In a second aspect, some embodiments of the present application further provide a 3D human ear model construction system, comprising: an acquisition module 100 , an extraction module 200 , a calculation module 300 and a construction module 400 .

[0103] The acquisition module 100 is used to acquire original point cloud data of the human ear.

[0104] The extraction module 200 is used to obtain human ear voxel features based on the original point cloud data and a preset first feature extraction algorithm; the extraction module 200 is also used to obtain human ear point cloud features based on the original point cloud data and a preset second feature extraction algorithm.

[0105] The calculation module 300 is used to obtain the bird's-eye view feature according to the human ear voxel feature and the preset mapping algorithm; the calculation module 300 is also used to integrate the human ear voxel feature, the bird's-eye view feature and the human ear point cloud feature with a preset integration algorithm to obtain a feature vector.

[0106] The construction module 400 is used to construct a three-dimensional model of the human ear according to the feature vector.

[0107] The earprint of the human ear is extracted by three-dimensional scanning to obtain the corresponding original point cloud data of the human ear. The original point cloud data is divided into two branches for processing. One branch extracts the features of the original point cloud data by a preset first feature extraction algorithm to obtain the voxel features of the human ear, and the other branch extracts the features of the original point cloud data by a preset second feature extraction algorithm to obtain the point cloud features of the human ear. The voxel features of the human ear are then mapped to a 2D feature bird's-eye view by a preset mapping algorithm to obtain the bird's-eye view features of the two-dimensional bird's-eye view. The voxel features of the human ear, the bird's-eye view features and the point cloud features of the human ear are then integrated by a preset integration algorithm to obtain a feature vector. A three-dimensional model of the human ear corresponding to the human ear is constructed according to the integrated feature vector. After the three-dimensional model of the human ear is constructed, the corresponding headphones can be designed based on the three-dimensional model to improve the comfort of the headphones.

[0108] The 3D human ear model construction system of the embodiment of the present application combines different point cloud processing methods to extract features, so that the final three-dimensional model of the human ear pays more attention to the tiny and complex structure of the human ear and is more compatible with the human ear. The headphones made based on the three-dimensional model are also more compatible with the human ear, thereby improving the comfort of the headphones.

[0109] The operation process of the 3D human ear model construction system of the present application is similar to the aforementioned 3D human ear model construction method. For the specific operation process, please refer to the aforementioned 3D human ear model construction method, which will not be repeated here.

[0110] In a third aspect, an embodiment of the present application also provides an electronic device.

[0111] In some embodiments, the electronic device includes: at least one processor, and a memory communicatively connected to the at least one processor; wherein the memory stores instructions, and the instructions are executed by the at least one processor, so that when the at least one processor executes the instructions, any one of the 3D human ear model construction methods in the embodiments of the present application is implemented.

[0112] The processor and the memory may be connected via a bus or other means.

[0113] The memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer executable programs, such as the 3D human ear model construction method described in the embodiment of the present application. The processor implements the above-mentioned 3D human ear model construction method by running the non-transitory software program and instructions stored in the memory.

[0114] The memory may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store and execute the above-mentioned 3D human ear model construction method. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0115] The non-transient software programs and instructions required to implement the above-mentioned 3D human ear model construction method are stored in the memory, and when executed by one or more processors, the 3D human ear model construction method mentioned in the above-mentioned first aspect embodiment is executed.

[0116] In a fourth aspect, embodiments of the present application also provide a computer-readable storage medium.

[0117] In some embodiments, a computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to execute the 3D human ear model construction method mentioned in the embodiment of the first aspect.

[0118] In some embodiments, the storage medium stores computer executable instructions, which are executed by one or more control processors, for example, by a processor in the above-mentioned electronic device, so that the above-mentioned one or more processors can execute the above-mentioned 3D human ear model construction method.

[0119] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0120] It will be appreciated by those skilled in the art that all or some of the steps and systems in the disclosed method above may be implemented as software, firmware, hardware and appropriate combinations thereof. Some physical components or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or a non-transitory medium) and a communication medium (or a temporary medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that may be used to store desired information and may be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0121] The embodiments of the present application are described in detail above in conjunction with the accompanying drawings, but the present application is not limited to the above embodiments. Various changes can be made within the knowledge of ordinary technicians in the relevant technical field without departing from the purpose of the present application. In addition, the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

Claims

1. A method for constructing a 3D human ear model, characterized in that: include: Get the original point cloud data of the human ear; Obtaining a human ear voxel feature according to the original point cloud data and a preset first feature extraction algorithm; Obtaining a human ear point cloud feature according to the original point cloud data and a preset second feature extraction algorithm; Obtaining a bird's-eye view feature according to the human ear voxel feature and a preset mapping algorithm; Integrate the human ear voxel feature, the bird's-eye view feature, and the human ear point cloud feature using a preset integration algorithm to obtain a feature vector; Constructing a three-dimensional model of a human ear according to the feature vector; The step of obtaining a human ear voxel feature according to the original point cloud data and a preset first feature extraction algorithm includes: Dividing the original point cloud data into voxels of equal proportions in space to obtain a plurality of human ear voxels; Performing voxel feature encoding on the plurality of human ear voxels to obtain a plurality of voxel-wise features; Perform 3D sparse convolution on the multiple voxel-wise features to obtain human ear voxel features.

2. The method for constructing a 3D human ear model according to claim 1, characterized in that: The human ear voxel includes: a first voxel and a second voxel, wherein the first voxel is characterized as a screened human ear voxel, and the second voxel is characterized as an unscreened human ear voxel; The 3D human ear model construction method further includes: The second voxel is screened according to a preset voxel interval to obtain the first voxel.

3. The method for constructing a 3D human ear model according to claim 1, characterized in that: The preset second feature extraction algorithm includes: a preset scale space extreme value retrieval algorithm, a preset T-net algorithm, a preset MLP network algorithm and a preset aggregation algorithm; The step of obtaining the human ear point cloud feature according to the original point cloud data and a preset second feature extraction algorithm includes: The original point cloud data is searched using the preset scale space extreme value search algorithm to obtain a spatial extreme value range interval; Adjusting the original point cloud data according to the spatial extreme value range, the preset T-net algorithm and the preset aggregation algorithm to obtain an adjustment vector; Feature extraction is performed according to the preset MLP network algorithm and the adjustment vector to obtain human ear point cloud features.

4. The method for constructing a 3D human ear model according to claim 3, characterized in that: The adjusting the original point cloud data according to the spatial extreme value range, the preset T-net algorithm and the preset aggregation algorithm to obtain an adjustment vector includes: Adjusting the original point cloud data using the preset T-net algorithm according to the spatial extreme value range to obtain adjusted point cloud data; The adjusted point cloud data is aggregated according to the preset aggregation algorithm to obtain aggregated point cloud data.

5. The method for constructing a 3D human ear model according to claim 1, characterized in that: The preset mapping algorithm is a downsampling algorithm.

6. The method for constructing a 3D human ear model according to claim 1, characterized in that: The step of integrating the human ear voxel feature, the bird's-eye view feature, and the human ear point cloud feature using a preset integration algorithm to obtain a feature vector includes: The human ear voxel feature, the bird's-eye view feature, and the human ear point cloud feature are sampled respectively to obtain a plurality of key point features; The multiple key point features are converted into feature space through a two-layer FCN network to obtain the feature vector.

7. A 3D human ear model construction system, characterized in that: include: An acquisition module, wherein the acquisition module is used to acquire original point cloud data of a human ear; An extraction module, the extraction module is used to obtain a human ear voxel feature according to the original point cloud data and a preset first feature extraction algorithm; The step of obtaining a human ear voxel feature according to the original point cloud data and a preset first feature extraction algorithm includes: Dividing the original point cloud data into voxels of equal proportions in space to obtain a plurality of human ear voxels; Performing voxel feature encoding on the plurality of human ear voxels to obtain a plurality of voxel-wise features; Performing 3D sparse convolution on the multiple voxel-wise features to obtain human ear voxel features; The extraction module is also used to obtain the human ear point cloud features according to the original point cloud data and a preset second feature extraction algorithm; A calculation module, the calculation module is used to obtain a bird's-eye view feature according to the human ear voxel feature and a preset mapping algorithm; The calculation module is further used to integrate the human ear voxel feature, the bird's-eye view feature, and the human ear point cloud feature using a preset integration algorithm to obtain a feature vector; A construction module is used to construct a three-dimensional model of a human ear according to the feature vector.

8. An electronic device, characterized in that The method comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the 3D human ear model construction method as described in any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the 3D human ear model construction method according to any one of claims 1 to 6.

Images