Methods and devices for processing 3D facial models, electronic devices, and storage media

Abstract

This application relates to a method, apparatus, electronic device, and storage medium for processing 3D facial models. The method includes: acquiring a 3D facial model to be processed after scanning a target face; if a target defect exists in the initial periorbital 3D model corresponding to the periorbital region in the 3D facial model to be processed, acquiring a periorbital region base corresponding to the periorbital region; and fitting the periorbital region base to the 3D facial model to be processed to obtain a target facial 3D model after defect repair. This application solves the technical problem of low efficiency in 3D human body modeling due to the need for manual adjustment of the scanned human body model.

Description

Technical Field

[0001] This application relates to the field of 3D modeling technology, and in particular to a method and apparatus for processing 3D facial models, electronic devices, and storage media. Background Technology

[0002] With the rapid development of computer technology and image processing technology, realistic 3D modeling technology is playing an increasingly important role in fields such as industrial design, film, games, and the internet. In human 3D modeling, the human body is typically scanned using a 3D human body scanner to generate a scanned model. However, due to limitations in scanning accuracy and various errors, this method often results in flaws in the generated scanned model. For example, facial scanned models may exhibit noticeable distortion in the eye area (eyelids, eye sockets), generally requiring manual adjustment. To reduce these flaws, related technologies for human 3D modeling often require manual adjustments to the scanned model, leading to lower accuracy and efficiency in the process.

[0003] There is currently no effective solution to the problem of low efficiency in human body 3D modeling due to the need for manual adjustments to the human body scan model. Summary of the Invention

[0004] This application provides a method and apparatus for processing 3D facial models, an electronic device, and a storage medium, to at least solve the technical problem in related technologies where the 3D modeling of the human body requires manual adjustment of the human body scanning model, leading to difficulties in 3D human body modeling.

[0005] According to one aspect of the embodiments of this application, a method for processing a three-dimensional face model is provided, comprising: acquiring a three-dimensional face model to be processed obtained after scanning a target face; in the case that there is a target defect in the initial three-dimensional model of the periorbital region corresponding to the periorbital region in the three-dimensional face model to be processed, acquiring a periorbital region base corresponding to the periorbital region; fitting the periorbital region base with the three-dimensional face model to be processed to obtain a target face three-dimensional model after defect repair.

[0006] According to another aspect of the embodiments of this application, a face 3D model processing device is also provided, including: a model acquisition module, used to acquire a face 3D model to be processed obtained after scanning a target face; a base acquisition module, used to acquire a base corresponding to the eye area in the initial eye area 3D model corresponding to the eye area in the face 3D model to be processed, when there is a target defect; and a model fitting module, used to fit the eye area base with the face 3D model to be processed to obtain a target face 3D model after defect repair.

[0007] According to another aspect of the embodiments of this application, a storage medium is also provided, the storage medium including a stored program that executes the above-described method when the program is run.

[0008] According to another aspect of the embodiments of this application, an electronic device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor performs the above-described method through the computer program.

[0009] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps of any of the embodiments of the methods described above.

[0010] In this embodiment, a three-dimensional model of the target face is obtained after scanning the target face. If the initial three-dimensional model of the periorbital region in the three-dimensional model of the target face has defects, a periorbital region base corresponding to the periorbital region is obtained. The periorbital region base is then fitted to the three-dimensional model of the target face to obtain a target three-dimensional model after defect repair. By obtaining the periorbital region base corresponding to the defective periorbital region in the scanned model, and fitting the periorbital region base to the defective scanned model, the target three-dimensional model of the target face is obtained. The periorbital region of the target three-dimensional model is free of defects, thus achieving automatic repair of defects in the scanned model and reducing defects in the human body three-dimensional model. This solves the technical problem in related technologies where the human body three-dimensional model requires manual adjustment of the human body scanned model, resulting in low work efficiency in human body three-dimensional modeling. Attached Figure Description

[0011] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0012] Figure 1 This is a schematic diagram of the hardware environment of the face 3D model processing method according to an embodiment of this application;

[0013] Figure 2 This is a flowchart of an optional method for processing a 3D face model according to an embodiment of this application;

[0014] Figure 3 This is a flowchart of another optional method for processing a 3D face model according to an embodiment of this application;

[0015] Figure 4 This is a schematic diagram of an optional three-dimensional face model processing device according to an embodiment of this application;

[0016] Figure 5 This is a schematic diagram of another optional face 3D model processing device according to an embodiment of this application; and,

[0017] Figure 6 This is a structural block diagram of a terminal according to an embodiment of this application. Detailed Implementation

[0018] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0019] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0020] First, some nouns or terms that appear in the description of the embodiments of this application shall be interpreted as follows:

[0021] A basis is a set of basis vectors in n-dimensional space. In n-dimensional space, the n basis vectors that form a basis must satisfy the following requirements: any vector in n-dimensional space can be represented as a linear combination of the n basis vectors, and the way this linear combination is represented must be unique (i.e., the n basis vectors need to satisfy the condition of linear independence).

[0022] Principal Component Analysis (PCA) is a mathematical dimensionality reduction method. The goal of PCA is to select k basis vectors as a new basis, reducing an n-dimensional vector to k dimensions. PCA transforms the original data into a set of linearly independent representations, also known as principal components, through linear transformation. These principal components can be used to extract the main feature components of the data, thus showcasing the data's characteristics in a smaller dimension. It is commonly used for dimensionality reduction of high-dimensional data.

[0023] According to one aspect of the embodiments of this application, a method embodiment for processing three-dimensional models is provided.

[0024] Optionally, in this embodiment, the above-described face 3D model processing method can be applied to, for example... Figure 1 The hardware environment shown consists of terminal 101 and server 103. Figure 1 As shown, server 103 is connected to terminal 101 via a network and can be used to provide services (such as 3D model processing services, model defect repair services, etc.) to the terminal or clients installed on the terminal. Database 105 can be set up on the server or independently of the server to provide data storage services for server 103. The aforementioned network includes, but is not limited to, wide area networks (WANs), metropolitan area networks (MANs), or local area networks (LANs). Terminal 101 is not limited to PCs, mobile phones, tablets, etc. The face 3D model processing method of this application embodiment can be executed by server 103, by terminal 101, or by both server 103 and terminal 101. Specifically, the face 3D model processing method of this application embodiment can also be executed by a client installed on terminal 101. The following description uses the execution of a face 3D model processing method of this application embodiment on a server as an example.

[0025] This application can be applied to scenarios including but not limited to human 3D modeling, animal 3D modeling, and object 3D modeling.

[0026] Figure 2 This is a flowchart of an optional face 3D model processing method according to an embodiment of this application, such as... Figure 2 As shown, the method may include the following steps:

[0027] Step S202: Obtain the 3D model of the face to be processed after scanning the target face;

[0028] Step S204: If there are target defects in the initial three-dimensional model of the eye area corresponding to the eye area in the three-dimensional model of the face to be processed, obtain the base of the eye area corresponding to the eye area.

[0029] Step S206: Fit the base of the eye area to the three-dimensional model of the face to be processed to obtain the three-dimensional model of the target face after the target blemish is repaired.

[0030] Through steps S202 to S206 above, by obtaining the base of the eye area corresponding to the flawed eye area in the scanned model, the base of the eye area is fitted to the flawed scanned model to obtain the target face 3D model. The eye area in the target face 3D model is flawless, realizing the automatic repair of flaws in the scanned model, achieving the purpose of reducing flaws in the human body 3D model, thereby solving the technical problem in related technologies that the human body 3D modeling requires manual adjustment of the human body scanned model, resulting in low work efficiency of human body 3D modeling.

[0031] In the technical solution provided in step S202, the server obtains the three-dimensional model of the face to be processed after scanning the target face.

[0032] The aforementioned 3D model of the face to be processed is a 3D model obtained based on the scanning information of the target face. It can be a 3D model directly generated after scanning the target face, or a 3D model obtained after processing the scanning information.

[0033] There are various ways to scan a target face, including but not limited to using a camera to capture images or using a scanner. Regardless of the method used, as long as information about each point on the target face can be obtained, a 3D model corresponding to the target face can be created based on the information about each point on the target face.

[0034] The target face mentioned above can be any type of face, and any object whose three-dimensional model can be built based on its three-dimensional contour can also be processed into a three-dimensional model according to the processing method in the embodiments of this application.

[0035] In the technical solution provided in step S204, if there are target defects in the initial three-dimensional model of the eye area corresponding to the eye area in the three-dimensional model of the face to be processed, the server obtains the base of the eye area corresponding to the eye area.

[0036] In order to automatically repair potential flaws around the eyes in the 3D facial model being processed, the server can determine whether there are flaws around the eyes in the 3D facial model being processed.

[0037] In this application, the determination of the existence of a target defect can be that the initial three-dimensional model corresponding to the eye area in the three-dimensional model of the face to be processed actually has a local three-dimensional model with a scanning error between the initial three-dimensional model corresponding to the eye area and the eye area of ​​the target face that is greater than or equal to a preset error. Alternatively, it can be determined based on preset conditions that the initial three-dimensional model corresponding to the eye area in the three-dimensional model of the face to be processed may have a local three-dimensional model with a scanning error between the initial three-dimensional model corresponding to the eye area and the target face that is greater than or equal to a preset error.

[0038] There are several ways to determine whether there are target defects in the initial three-dimensional model of the eye area corresponding to the eye area in the three-dimensional model of the face to be processed, including but not limited to the following: (1) Obtain the surface smoothness of the initial three-dimensional model of the eye area of ​​the three-dimensional model of the face to be processed and the average surface smoothness of the face of the target face type in the eye area. If the difference between the surface smoothness of the initial three-dimensional model of the eye area and the average surface smoothness is greater than or equal to the preset error smoothness, it is confirmed that the scanning error between the initial three-dimensional model of the eye area and the target face is greater than or equal to the preset error, and it is determined that there are target defects; (2) According to the scanning modeling experience of the face of the target face type (i.e. the face type of the target face), it is known that the eye area of ​​the face of the target face type is prone to defects. Therefore, the eye area is regarded as a preset part of the target face type where there are target defects. After obtaining the three-dimensional model of the face to be processed, the server determines that there are target defects in the initial three-dimensional model of the eye area corresponding to the eye area in the three-dimensional model of the face to be processed according to the face type corresponding to the three-dimensional model of the face to be processed. (3) Sample feature points of the initial three-dimensional model around the eyes of the face to be processed. Collect the same number of feature points of the target face in this area. The number of feature points is significantly greater than the number of coordinate points used in this area when the three-dimensional model of the face to be processed is established. Fit the feature points collected from the initial three-dimensional model around the eyes and the target face respectively. If the number of feature points that cannot be fitted to the same surface is greater than or equal to the preset number, it is determined that the initial three-dimensional model around the eyes has target defects.

[0039] The aforementioned eye periorbital area refers to the area around the eyes within the face of the target face type. The target face type is the face type of the target face. For example, if the target face is the face of an Asian young male A, the target face type is an Asian young male face, and the eye periorbital area refers to the area around the eyes within the face of an Asian young male.

[0040] After the server determines that there are target defects in the 3D model of the face to be processed (i.e., defects around the eyes need to be repaired), in order to automatically repair the defects in the initial 3D model of the eye area corresponding to the eye area in the 3D model of the face to be processed, the server obtains the base of the eye area corresponding to the eye area.

[0041] The aforementioned periorbital region base is obtained by training multiple periorbital region samples. This base is capable of reconstructing a 3D model corresponding to the periorbital region according to a certain coefficient. Each periorbital region sample is obtained by sampling the periorbital region of a target face type. Since the periorbital region samples are free of defects, the 3D model reconstructed from the periorbital region base trained on these samples will also be free of defects. The periorbital region base, obtained after a single training iteration, can be used multiple times without requiring retraining each time the method of this application is implemented.

[0042] The periorbital base can be obtained by means including but not limited to the following methods: (1) training multiple periorbital samples using Principal Component Analysis (PCA) to obtain the periorbital base; (2) training multiple periorbital samples using Linear Discriminant Analysis (LDA) to obtain the periorbital base.

[0043] Both PCA and LDA algorithms can be used for dimensionality reduction to obtain the basis. Both can reduce the dimensionality of data and both use the idea of ​​matrix eigenvalue decomposition. Both assume that the data follows a Gaussian distribution. However, LDA is a supervised dimensionality reduction method, while PCA is an unsupervised one. LDA can reduce dimensionality to a maximum of k-1 dimensionality, while PCA has no such limitation. In addition to dimensionality reduction, LDA can also be used for classification. LDA selects the projection direction with the best classification performance, while PCA selects the direction with the maximum variance of the sample point projection.

[0044] The periocular base can be obtained by training multiple samples from a single periocular region, or by training multiple sets of samples from different regions as a group of training samples. Each group of training samples comes from the same 3D model. The periocular samples are included among the samples from multiple different regions. When multiple samples from different regions are used as a group of training samples, the target base obtained includes both the periocular base representing the periocular region and the base representing other regions. The periocular base can be extracted from the target base. The target base, obtained after one training, can be used multiple times, or it can be split into multiple region bases (such as a periocular base and additional region bases, where the additional region bases can be an ocular base, a lacrimal gland base, an eyelash base, etc.) for separate use, without needing to retrain each time the method of this application is implemented.

[0045] For example, if the target face is A, and A is an Asian young male, then the target face type is an Asian young male face. 100 head samples of Asian young men are obtained through manual creation or online collection. The periorbital area is cropped from each head sample to obtain periorbital area samples (the samples are flawless). Principal component analysis is used to train these 100 periorbital area samples to obtain an orthogonal basis, namely the periorbital area basis. The periorbital area basis has a strong representational ability and can reconstruct a three-dimensional model of the periorbital area (of an Asian young male) through a certain coefficient.

[0046] There are several ways to obtain samples of the periorbital area. For example, a 3D model of the periorbital area can be cut out from a batch of flawless human head models; another way is to obtain a dataset of 3D models of the periorbital area of ​​the human body from a human 3D model database.

[0047] In the technical solution provided in step S206, the server fits the base of the periorbital area with the three-dimensional model of the face to be processed, and obtains the three-dimensional model of the target face after the target blemish is repaired.

[0048] Optionally, there are several ways to fit the periorbital 3D model to the face to be processed. For example, the initial periorbital 3D model can be replaced with a periorbital model reconstructed according to the target fitting coefficients to obtain the target face 3D model; another example is to adjust the initial periorbital 3D model in the face to be processed according to the periorbital model so that the periorbital area of ​​the adjusted target face 3D model does not contain the target defects. Specifically, the periorbital 3D model can be fitted to the face to be processed 3D model in the following ways:

[0049] (1) Find a target fitting coefficient so that the eye periphery model reconstructed according to the target fitting coefficient can fit the eye periphery in the three-dimensional face model to be processed as closely as possible. The eye periphery model reconstructed according to the target fitting coefficient is located at the position of the eye periphery in the three-dimensional face model to be processed. It is fused with the three-dimensional face model to be processed to obtain the target three-dimensional face model. If the fusion edge is not smooth, the fusion edge can be smoothed.

[0050] Optionally, there are several ways to obtain the target fitting coefficient: ① Calculate the distance error between the 3D model reconstructed according to a certain fitting coefficient and the corresponding points in the 3D model of the periorbital region, and then correct the fitting coefficient until the error meets the requirements or the maximum number of iterations is met. Take the last fitting coefficient as the target fitting coefficient, and match the periorbital region model reconstructed according to the target fitting coefficient with the periorbital region in the 3D model of the face to be processed. ② Use the gradient descent algorithm to obtain the target fitting coefficient.

[0051] (2) A deep learning network is used to fit the basal part of the eye area to the three-dimensional model of the face to be processed, so as to obtain the fitted target three-dimensional model of the face.

[0052] Since the periorbital base is obtained by training with samples of the periorbital area without defects, the target face 3D model obtained by fitting the periorbital base and the 3D face model to be processed also has no defects in the periorbital area.

[0053] As an optional embodiment, step S202, obtaining the 3D model of the face to be processed obtained after scanning the target face, further includes the following steps:

[0054] Step S2021: Obtain the initial 3D model after scanning the target face;

[0055] Step S2022: Align the initial 3D model with the preset template to obtain the 3D model of the face to be processed, wherein the preset template is a preset 3D model of the target face type and the target face type is the face type of the target face.

[0056] Alignment methods include, but are not limited to: ICP (Iterative Closest Points) algorithm, PL-ICP (point-line Iterative Closest Points) algorithm, and NICP (Normal Iterative ClosestPoint) algorithm.

[0057] To unify the topology, the initial 3D model needs to be aligned to a preset template. The preset template can be a meshed 3D model of the target face type. By aligning, the preset 3D model can be fitted to the shape of the initial 3D model, resulting in a meshed 3D model of the face to be processed.

[0058] The aforementioned initial 3D model can be a model reconstructed from the target face through multi-viewpoint methods or depth maps, or it can be a scanned model generated by scanning the target face with a scanner.

[0059] Optionally, in this embodiment, step S2021, obtaining the initial 3D model after scanning the target face, further includes the following steps:

[0060] Step S11: Obtain a multi-view 2D image of the target face;

[0061] Step S12: Using the multi-viewpoint 2D images and viewpoint information from each viewpoint, reconstruct the initial 3D model of the target face.

[0062] Optionally, in this embodiment, step S2021, obtaining the initial 3D model after scanning the target face, further includes the following steps:

[0063] Step S21: Obtain the depth map of the target face, wherein the depth map of the target face is used to represent the distance between each point on the target face and the camera on a two-dimensional plane;

[0064] Step S22: Using the depth map of the target face, reconstruct the initial three-dimensional model of the target face.

[0065] As an optional embodiment, step S204, obtaining the periorbital base corresponding to the periorbital region, may further include the following steps:

[0066] Step S61: Obtain multiple sets of training samples, wherein each set of training samples includes at least a sample of the periorbital area, which is a three-dimensional model of the periorbital area and has no defects.

[0067] Step S62: Analyze multiple training samples using principal component analysis to obtain the periorbital basal region.

[0068] Each training sample includes samples of the periorbital area, and may also include samples of other areas. For example, samples of the periorbital area, eyeball, and lacrimal gland can be obtained from a sample 3D model as a set of training samples. Multiple sets of training samples can be obtained through multiple sample 3D models, and the periorbital area in each sample 3D model is free of defects.

[0069] Analyzing multiple training samples using Principal Component Analysis (PCA) involves representing each training sample as a vector (concatenating the vectors of samples from the periorbital region and other regions into a single vector). PCA is then used to model the multiple vectors corresponding to these training samples, resulting in a new basis vector. Following the method used to concatenate the vectors of the periorbital region and other regions during training sample construction, this new basis vector is then split, yielding the periorbital basis vector and the basis vectors for other regions.

[0070] As an optional embodiment, step S206, which involves fitting the periorbital area base to the three-dimensional model of the face to be processed to obtain a three-dimensional model of the target face after blemish repair, further includes the following steps:

[0071] Step S81: The specified three-dimensional model that is reconstructed from the base of the periorbital region according to the target fitting coefficient and meets the preset fitting conditions is determined as the periorbital region model. The preset fitting conditions are that the fitting degree between the initial periorbital three-dimensional model and the specified three-dimensional model is greater than or equal to the preset fitting degree.

[0072] Step S82: Fuse the periorbital model with the 3D model of the face to be processed to obtain the target 3D model of the face.

[0073] The aforementioned initial 3D model of the periorbital region is the initial 3D model of the periorbital area containing the target blemishes in the 3D model of the face to be processed. The periorbital region model is obtained by dimensional reconstruction based on the periorbital region basal layer and is used as a new 3D model of the periorbital region in the 3D model of the face to be processed.

[0074] Because additional parts cannot be accurately scanned during scanning and modeling, or because they are obscured by the periorbital area, it is necessary to generate a separate 3D model for the additional parts around the eyes to make the 3D model of the target face more complete and accurate in presenting the structure of the target face.

[0075] As an optional embodiment, in step S206, after fitting the periorbital area base with the three-dimensional model of the face to be processed to obtain the target face three-dimensional model after blemish repair, the method further includes the following steps:

[0076] Step S302: Obtain the base of the additional part corresponding to the additional part, wherein the additional part is the part in the face of the target face type that is related to the periorbital area, and the target face type is the face type of the target face.

[0077] Step S304: Reconstruct the additional part base into an additional part model according to the target fitting coefficient, and assemble the additional part model into the target face 3D model.

[0078] The aforementioned relationships can refer to connections, occlusions, or inclusions between the additional part and the periorbital area within the target face type. For example, the additional part could be the eyeball, lacrimal gland, or eyelashes.

[0079] The acquisition method of the additional part base is similar to that of the periorbital part base, that is, it is obtained by training multiple additional part samples. However, if the additional part base and the periorbital part base are trained separately, there may be a mismatch or lack of fit between the additional part model (reconstructed from the additional part base) and the periorbital part model (reconstructed from the periorbital part base). It may be necessary to perform a second fitting. In order to improve the efficiency of model processing and the accuracy of generating additional parts, it is necessary to train the additional part base and the periorbital part base at the same time. The additional part base and the periorbital part base obtained by training at the same time can be reconstructed into a matching and fit additional part model and periorbital part model through the same coefficients.

[0080] Since the additional feature base and the periorbital feature base obtained through simultaneous training can be reconstructed using the same coefficients, the target fitting coefficients used for reconstruction can be fitted using either the periorbital feature base or the additional feature base. The choice of method depends on the specific circumstances. For cases where the additional feature is the eyeball, the target fitting coefficients are typically obtained by fitting the periorbital feature base to the 3D face model. Compared to the eyeball, the periorbital region exhibits greater variation across different face models. For instance, a person with a rounder face and a person with a thinner face may have very similar eyeball sizes and shapes, but their periorbital regions will show significant differences. The rounder face may have a wider periorbital region. Therefore, using the periorbital feature base for fitting will yield more accurate coefficients.

[0081] Optionally, before obtaining the additional part base corresponding to the additional part in step S302, the method further includes the following steps:

[0082] Step S71: Obtain multiple sets of training samples, wherein each set of training samples includes at least a periorbital sample and an additional sample. The periorbital sample is a three-dimensional model of the periorbital area obtained from the sample three-dimensional model of the target face type. There are no flaws in the periorbital sample. The additional sample is a three-dimensional model corresponding to the additional part obtained from the sample three-dimensional model.

[0083] Step S72: Analyze multiple training samples using principal component analysis to obtain the target basis;

[0084] Step S73: Extract the periorbital base and additional base from the target base.

[0085] Steps S71 to S73 described above correspond to steps S61 to S62 in another embodiment. Their purpose is to obtain the base through principal component analysis of the training samples. When it is necessary to generate a three-dimensional model of the additional parts, the training of the base of the periorbital region is completed at the same time as the training of the base of the periorbital region. In addition, the additional part samples and periorbital region samples in each training sample come from the same sample three-dimensional model. That is to say, in each training sample, the additional part samples and periorbital region samples are matched and fit together. If the periorbital region base and the additional part base obtained by training with such multiple sets of training samples are reconstructed using the same coefficients, the resulting periorbital region model and the additional part model will also be matched and fit together.

[0086] There can be multiple additional areas around the eyes. When there are multiple additional areas, the training can be carried out in the manner described in steps S71 to S73 above.

[0087] For example, if the target face is the face of an adult Asian male, then a 3D model of an adult Asian male needs to be obtained as a sample 3D model. The additional parts are the eyeball and the lacrimal gland. Then, from each sample 3D model, the periorbital sample is cropped, and the eyeball and lacrimal gland samples are extracted to obtain a set of training samples. The principal component analysis method is used to analyze multiple sets of training samples to obtain the target base, which includes the periorbital base, the eyeball base, and the lacrimal gland base.

[0088] Optionally, step S71, obtaining multiple sets of training samples, further includes the following steps:

[0089] Step S711: Obtain multiple sample 3D models of the target face type, wherein the sample 3D models have no flaws in the area around the eyes;

[0090] Step S712: Using multiple sample 3D models, obtain multiple sets of training samples, including obtaining each set of training samples according to the following steps:

[0091] Step S713: For any sample 3D model, cut out the periorbital area sample from the sample 3D model, and extract additional area samples from the sample 3D model to obtain a set of training samples.

[0092] Steps S711 to S713 above are for extracting samples of each part of the training substrate from the sample 3D model.

[0093] Optionally, step S72, analyzing multiple sets of training samples using principal component analysis to obtain the target basis, further includes the following steps:

[0094] S721, For each training sample, feature extraction is performed on the samples around the eyes to obtain a first feature vector, and feature extraction is performed on the samples of the additional areas to obtain a second feature vector. The first feature vector and the second feature vector are concatenated to form the target feature vector of the training sample.

[0095] S722, the target eigenvectors are analyzed using principal component analysis to obtain the target basis.

[0096] For each set of training samples, the model needs to be vectorized before training. There are various ways to vectorize the model, which can be determined according to the requirements.

[0097] For example, if the target face is an Asian adult male, then a 3D model of the Asian adult male needs to be obtained as a sample 3D model, with additional parts being the eyeball and lacrimal gland. From each sample 3D model, samples of the periorbital region are cropped, and samples of the eyeball and lacrimal gland are extracted to obtain a set of training samples. The coordinates of 3000 points in the periorbital region sample, 2000 points in the eyeball sample, and 1000 points in the lacrimal gland sample in a preset coordinate system are obtained. The preset coordinate system can be the coordinate system of the 3D model of the face to be processed. These coordinates are concatenated in order, and the feature matrix formed by the coordinates of 6000 points in the preset coordinate system is used as the target feature vector of this set of training samples. The multiple target feature vectors are analyzed using principal component analysis to obtain the target basis. The feature matrix of the target basis has a dimension of 6000×k, where k is the feature dimension after dimensionality reduction using principal component analysis. From the target basis, the periorbital region base, eyeball base, and lacrimal gland base can be separated in the concatenation order.

[0098] As an optional embodiment, the technical solution of this application is illustrated below with reference to specific embodiments:

[0099] Realistic human head models typically obtain geometric data through rawscan reconstruction (initial scan model), followed by retopology to obtain the model used for rigging. However, common alignment methods (such as the NICP algorithm) often result in noticeable distortions in the eyes (eyelids, eye sockets), requiring manual adjustments. Furthermore, matching components like the eyeballs need to be manually fabricated and placed, which is time-consuming. This solution aims to improve the efficiency of regularizing the eye shape in a 3D human head mesh and assembling related components by combining PCA models to automatically regularize flawed alignment models and assemble corresponding components.

[0100] Models that directly align using the NICP algorithm through keypoints generally retain the details of the scanned face, but are often limited by scanning accuracy and various errors, resulting in some imperfections (e.g., irregular eye sockets in directly aligned faces, or misalignment of corresponding components when placed directly using templates). This method first uses a well-formed head model (manually created or collected online) and pre-assembled components as training data to train a PCA model (training only once is sufficient if the training data is not modified). Each set of data contains the face shape and corresponding components (e.g., the eye mesh). After aligning to obtain an imperfect mesh, we use the corresponding PCA algorithm containing the basal part of the face mesh for fitting. Applying the fitted coefficients to the corresponding components yields the corresponding face components.

[0101] like Figure 3As shown, the overall implementation method, taking eye-related components as an example, is as follows: First, a batch of flawless models with face components are obtained through methods such as manual creation or online collection. The periocular region is cropped out (i.e., samples of the periocular region are obtained from the sample 3D model), and corresponding components such as the eyeball mesh and lacrimal gland mesh are extracted (i.e., additional region samples are obtained from the sample 3D model). The mutually matched periocular region, eyeball, lacrimal gland, and other meshes (the eyeball and lacrimal gland may or may not be included in the modeling depending on the situation; for example, eyelashes may be added or not) constitute a set of training data. The training data is vectorized and PCA modeling is performed (PCA modeling is for subsequent fitting of the periocular mesh coefficients; PCA modeling can obtain a set of PCA basis; other modeling methods can be used, and the obtained basis can be orthogonal or non-orthogonal basis). When the scanned model is obtained (which can be obtained by collecting and reconstructing from real faces through multi-viewpoint methods or depth maps), model alignment is performed first, aligning the scanned model to the template mesh to unify the topology. Then, the aligned, flawed mesh K is fitted to the base representing the periorbital region in the PCA base to obtain the fitting coefficient P (the fitted coefficient is used to increase the dimension of the base to obtain the reconstructed mesh, and the distance error of the corresponding point is calculated with the corresponding region of the deformed face obtained by NICP alignment (other loss functions can also be used), and then the fitting coefficient is corrected until the error meets the requirements or the maximum number of iterations is met); the fitting coefficient P is used to reconstruct the components of the base representing other components in the PCA base to obtain the corresponding and fitting eyeball and other component mesh (that is, the fitting coefficient P is used to increase the dimension of the base of other components to obtain the reconstructed mesh); the periorbital region reconstructed by the PCA base is fused into the aligned, flawed mesh K, the fusion edge is smoothed, and other components are placed in the reconstructed position to obtain a regular eye shape and corresponding components (other components reconstructed according to the fitting coefficient are directly located in the corresponding position on the mesh K).

[0102] The above overall implementation method can be achieved through, for example... Figure 4 The apparatus shown is complete. The scan model alignment module is used to align the scan model onto the template mesh (i.e., as shown in the diagram). Figure 3 The "wrap face alignment" module (shown) unifies the topological structure; the training data vectorization module is used to vectorize the training data; the PCA basis modeling module is used to obtain a set of PCA basis vectors through PCA modeling (i.e., such as...). Figure 3 The "base" shown includes a PCA base representing the periocular region and bases representing other components; the PCA base fitting module is used to fit the aligned, flawed mesh K with the base representing the periocular region in the PCA base to obtain the fitting coefficient P; the alignment model fusion module is used to fuse the periocular region reconstructed from the PCA base (i.e., Figure 3The "periocular component mesh" is merged into the imperfect mesh K after alignment; the eyeball and other component reconstruction module is used to reconstruct the components of the PCA substrate representing other components using the fitting coefficient P, to obtain the corresponding and fitting eyeball and other component mesh (i.e. Figure 3 The "eyeball mesh" and "tear gland mesh" are placed at the reconstruction location; the post-processing module is used to perform operations such as smoothing the fusion edges; the final human head 3D model is flawless around the eyes and contains eyeball and tear gland components.

[0103] In this embodiment, the PCA substrate is the target substrate in this application. The substrate representing the periorbital region in the PCA substrate is the periorbital part substrate in this application. The substrate representing other components in the PCA substrate is the additional part substrate in this application. The fitting coefficient P is the target fitting coefficient in this application. The flawed mesh K is the three-dimensional face model to be processed in this application. The three-dimensional human head model is the target three-dimensional face model in this application.

[0104] This solution combines a PCA base, providing certain data constraints and maintaining a good mesh shape. It also leverages the ability of bases within the same data set to fit together, automatically generating the corresponding eye components. This solution can be further applied to the generation of oral cavity components (such as teeth and tongue). This approach improves the efficiency of 3D digital human production and reduces labor costs.

[0105] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0106] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0107] According to another aspect of the embodiments of this application, a face three-dimensional model processing apparatus for implementing the above-described face three-dimensional model processing method is also provided. Figure 5 This is a schematic diagram of another optional three-dimensional face model processing device according to an embodiment of this application, such as... Figure 5 As shown, the device may include:

[0108] Model acquisition module 22 is used to acquire the 3D model of the face to be processed obtained after scanning the target face;

[0109] The base acquisition module 24 is used to acquire the base of the eye periphery corresponding to the eye periphery in the initial eye periphery 3D model corresponding to the eye periphery in the face 3D model to be processed when there are target defects.

[0110] The model fitting module 26 is used to fit the base of the periorbital area with the three-dimensional model of the face to be processed, so as to obtain the three-dimensional model of the target face after the target blemish is repaired.

[0111] It should be noted that the model acquisition module 22 in this embodiment can be used to execute step S202 in this application embodiment, the basis acquisition module 24 in this embodiment can be used to execute step S204 in this application embodiment, and the model fitting module 26 in this embodiment can be used to execute step S206 in this application embodiment.

[0112] It should be noted that the examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiments. It should also be noted that the above modules, as part of a device, can operate in environments such as... Figure 1 The hardware environment shown can be implemented either through software or through hardware.

[0113] Through the above modules, by obtaining the base of the eye area corresponding to the flawed eye area in the scanned model, the base of the eye area is fitted to the flawed scanned model to obtain the target face 3D model. The eye area in the target face 3D model is flawless, realizing the automatic repair of flaws in the scanned model, achieving the goal of reducing flaws in the human body 3D model, thereby solving the technical problem of low work efficiency in human body 3D modeling due to the need for manual adjustment of the human body scanned model in related technologies.

[0114] As an optional embodiment, the base acquisition module 24 includes a base modeling unit, used to: acquire multiple sets of training samples before acquiring the base of the periocular region corresponding to the three-dimensional model of the periocular region, wherein each set of training samples includes at least a periocular region sample, the periocular region sample is a three-dimensional model of the periocular region, and there are no defects in the periocular region sample; and analyze the multiple sets of training samples according to the principal component analysis method to obtain the periocular region base of the periocular region.

[0115] As an optional embodiment, the model acquisition module 22 includes a model alignment unit, used to: acquire an initial three-dimensional model obtained after scanning the target face; align the three-dimensional model of the face to be processed onto a preset template to obtain a three-dimensional model of the face to be processed, wherein the preset template is a preset three-dimensional model of the target face type, and the target face type is the face type of the target face.

[0116] As an optional embodiment, the model fitting module 26 includes a base fitting unit, used to: determine a specified three-dimensional model that is reconstructed from the base of the periorbital region according to the target fitting coefficient and meets the preset fitting conditions as the periorbital region model, wherein the preset fitting conditions are that the fitting degree between the initial periorbital three-dimensional model and the specified three-dimensional model is greater than or equal to the preset fitting degree; and a model fusion unit, used to fuse the periorbital region model with the three-dimensional model of the face to be processed to obtain the target three-dimensional model of the face.

[0117] Optionally, the base acquisition module 24 further includes an additional component base acquisition unit, used to: after fusing the periorbital region model with the three-dimensional face model to be processed, acquire the additional component base corresponding to the additional component, wherein the additional component is a part in the face of the target face type that has a relationship with the periorbital region, and the target face type is the face type of the target face.

[0118] Optionally, the model fitting module 26 further includes an additional component base reconstruction unit, used to: reconstruct the additional component base into an additional component model according to the target fitting coefficient, and assemble the additional component model into the target face 3D model.

[0119] Optionally, the basal modeling unit is further configured to: acquire multiple sets of training samples, wherein each set of training samples includes at least periorbital region samples and additional region samples, wherein the periorbital region samples are three-dimensional models of the periorbital region obtained from the sample three-dimensional model of the target face type, and the periorbital region samples are free of defects, and the additional region samples are three-dimensional models corresponding to the additional regions obtained from the sample three-dimensional model; analyze the multiple sets of training samples according to the principal component analysis method to obtain the target basal; and extract the periorbital region basal and the additional region basal from the target basal.

[0120] Optionally, the basal modeling unit is also used to: for each group of training samples, extract features from samples around the eye to obtain a first feature vector, and extract features from samples in additional areas to obtain a second feature vector, concatenate the first feature vector and the second feature vector to form the target feature vector of the training samples; and analyze the target feature vector according to the principal component analysis method to obtain the target basal.

[0121] Optionally, the basal modeling unit is also used to: acquire multiple sample 3D models of the target face type, wherein the sample 3D models have no defects in the peri-eye area; and acquire multiple sets of training samples using the multiple sample 3D models, including acquiring each set of training samples according to the following steps: for any sample 3D model, cut out the target part sample from the sample 3D model according to the peri-eye area, and extract additional part samples from the sample 3D model to obtain a set of training samples.

[0122] It should be noted that the examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiments. It should also be noted that the above modules, as part of a device, can operate in environments such as... Figure 1 The hardware environment shown can be implemented through software or hardware, and the hardware environment includes the network environment.

[0123] According to another aspect of the embodiments of this application, a server or terminal for implementing the above-described face 3D model processing method is also provided.

[0124] Figure 6 This is a structural block diagram of a terminal according to an embodiment of this application, such as... Figure 6 As shown, the terminal may include: one or more ( Figure 6 Only one of the following is shown: processor 601, memory 603, and transmission device 605, as shown in the image. Figure 6 As shown, the terminal may also include input / output devices 607.

[0125] The memory 603 can be used to store software programs and modules, such as the program instructions / modules corresponding to the face 3D model processing method and apparatus in this embodiment. The processor 601 executes various functional applications and data processing by running the software programs and modules stored in the memory 603, thereby realizing the aforementioned face 3D model processing method. The memory 603 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 603 may further include memory remotely located relative to the processor 601, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0126] The aforementioned transmission device 605 is used to receive or send data via a network, and can also be used for data transfer between the processor and memory. Specific examples of the network described above may include wired networks and wireless networks. In one example, the transmission device 605 includes a Network Interface Controller (NIC), which can be connected to other network devices and a router via a network cable to communicate with the Internet or a local area network. In another example, the transmission device 605 is a radio frequency (RF) module used for wireless communication with the Internet.

[0127] Specifically, memory 603 is used to store application programs.

[0128] The processor 601 can call the application stored in the memory 603 through the transmission device 605 to perform the following steps: acquiring a three-dimensional model of the face to be processed obtained after scanning the target face; if there are target defects in the initial three-dimensional model of the periorbital region corresponding to the periorbital region in the three-dimensional model of the face to be processed, acquiring the periorbital region base corresponding to the periorbital region; fitting the periorbital region base with the three-dimensional model of the face to be processed to obtain a three-dimensional model of the target face after defect repair.

[0129] This application provides a solution for processing three-dimensional models. By obtaining the base of the periorbital region corresponding to the flawed periorbital region in the scanned model, the base of the periorbital region is fitted to the flawed scanned model to obtain a target three-dimensional face model. The periorbital region in the target three-dimensional face model is flawless, thus achieving automatic repair of flaws in the scanned model and reducing flaws in the human body three-dimensional model. This solves the technical problem in related technologies where the human body three-dimensional modeling requires manual adjustment of the human body scanned model, resulting in low work efficiency. In addition, since it is impossible to accurately scan the additional parts of the periorbital region during scanning modeling, or the additional parts cannot be scanned because they are occluded by the periorbital region, this solution also automatically generates a three-dimensional model for the additional parts of the periorbital region, which is used to assemble into the target three-dimensional face model, making the target three-dimensional face model present the structure of the target face more completely and accurately.

[0130] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments, and will not be repeated here.

[0131] Those skilled in the art will understand that Figure 6 The structure shown is for illustrative purposes only. The terminal can be a smartphone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile internet device (MID), a PAD, or other terminal devices. Figure 6 This does not limit the structure of the aforementioned electronic devices. For example, the terminal may also include components that are more... Figure 6 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 6 The different configurations shown.

[0132] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0133] Embodiments of this application also provide a storage medium. Optionally, in this embodiment, the storage medium can be used to execute program code for a face 3D model processing method.

[0134] Optionally, in this embodiment, the storage medium may be located on at least one of the network devices in the network shown in the above embodiment.

[0135] Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps:

[0136] S202, Obtain the 3D model of the target face after scanning the target face;

[0137] S204, If there are target defects in the initial three-dimensional model of the eye area corresponding to the eye area in the three-dimensional model of the face to be processed, obtain the base of the eye area corresponding to the eye area.

[0138] S206, Fit the base of the eye area to the three-dimensional model of the face to be processed to obtain the three-dimensional model of the target face after the target blemish is repaired.

[0139] Optionally, the storage medium is also configured to store program code for performing the following steps: obtaining a three-dimensional model of the face to be processed after scanning the target face; if there are target defects in the initial three-dimensional model of the periorbital region corresponding to the periorbital region in the three-dimensional model of the face to be processed, obtaining the periorbital region base corresponding to the periorbital region; fitting the periorbital region base to the three-dimensional model of the face to be processed to obtain a three-dimensional model of the target face after defect repair.

[0140] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments, and will not be repeated here.

[0141] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0142] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0143] If the integrated units in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in the aforementioned computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause one or more computer devices (which may be personal computers, servers, or network devices, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.

[0144] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0145] In the several embodiments provided in this application, it should be understood that the disclosed client can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, indirect coupling or communication connection between units or modules, and may be electrical or other forms.

[0146] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0147] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0148] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for processing 3D facial models, characterized in that, include: Obtain the 3D model of the target face after scanning the target face; Determine whether there is a target defect in the initial periorbital 3D model corresponding to the periorbital region in the 3D model of the face to be processed. The determination includes: obtaining the surface smoothness of the initial periorbital 3D model and the average surface smoothness of the target face type. If the difference between the two is greater than or equal to a preset error smoothness, it is determined that there is a target defect. If there is a target defect in the initial three-dimensional model of the eye area corresponding to the eye area in the three-dimensional model of the face to be processed, the base of the eye area corresponding to the eye area is obtained. The target defect refers to a local three-dimensional model in the initial three-dimensional model of the eye area where the scanning error between the eye area and the target face is greater than or equal to a preset error. Fitting the periorbital area base to the 3D model of the face to be processed yields a 3D model of the target face after flaw repair, including: Obtain the additional site base corresponding to at least one additional site that is related to the periorbital region, wherein the periorbital region base and the additional site base are extracted from the target base obtained by performing principal component analysis on multiple sets of training samples from the same batch of sample 3D models. Each set of training samples includes periorbital region samples and additional site samples from the same sample 3D model. The periorbital region base is reconstructed into a periorbital region model according to the target fitting coefficient, and the additional region base is reconstructed into an additional region model using the target fitting coefficient; The periorbital region model and the additional region model are fused with the 3D face model to be processed, and the fused edges are smoothed to obtain the target 3D face model.

2. The method according to claim 1, characterized in that, Before obtaining the substrate of the additional part corresponding to the additional part, the method further includes: Multiple sets of training samples are obtained, wherein each set of training samples includes at least a periorbital sample and an additional sample. The periorbital sample is a three-dimensional model of the periorbital region obtained from the sample three-dimensional model of the target face type. The periorbital sample is free of defects. The additional sample is a three-dimensional model corresponding to the additional region obtained from the sample three-dimensional model. The target basis is obtained by analyzing the multiple sets of training samples using principal component analysis. The periorbital region base and the additional region base are extracted from the target base.

3. The method according to claim 2, characterized in that, The step of analyzing the multiple training samples using principal component analysis to obtain the target basis includes: For each set of training samples, feature extraction is performed on the periorbital area samples to obtain a first feature vector, and feature extraction is performed on the additional area samples to obtain a second feature vector. The first feature vector and the second feature vector are concatenated to form the target feature vector of the training sample. The target eigenvectors are analyzed using principal component analysis to obtain the target basis.

4. The method according to claim 1, characterized in that, Before obtaining the periorbital base corresponding to the periorbital region, the method further includes: Multiple sets of training samples are obtained, wherein each set of training samples includes at least a sample of the periorbital area, the periorbital area sample is a three-dimensional model of the periorbital area, and the periorbital area sample is free of defects; The multiple training samples were analyzed using principal component analysis to obtain the periorbital base corresponding to the periorbital region.

5. The method according to claim 1, characterized in that, The process of obtaining the 3D model of the target face after scanning the target face includes: Obtain the initial 3D model after scanning the target face; The three-dimensional model of the face to be processed is aligned with a preset template to obtain the three-dimensional model of the face to be processed, wherein the preset template is a preset three-dimensional model of the target face type, and the target face type is the face type of the target face.

6. A facial 3D model processing device, characterized in that, include: The model acquisition module is used to acquire the 3D model of the target face after scanning the target face. The substrate acquisition module is used to determine whether there is a target defect in the initial periorbital 3D model corresponding to the periorbital region in the 3D model of the face to be processed. The determination includes: acquiring the surface smoothness of the initial periorbital 3D model and the average surface smoothness of the target face type. If the difference between the two is greater than or equal to a preset error smoothness, it is determined that there is a target defect. If there is a target defect in the initial periorbital 3D model corresponding to the periorbital region in the 3D model of the face to be processed, the substrate of the periorbital region corresponding to the periorbital region is acquired. The model fitting module is used to fit the periorbital area base to the three-dimensional model of the face to be processed, to obtain a three-dimensional model of the target face after flaw repair, including: Obtain the additional site base corresponding to at least one additional site that is related to the periorbital region, wherein the periorbital region base and the additional site base are extracted from the target base obtained by performing principal component analysis on multiple sets of training samples from the same batch of sample 3D models. Each set of training samples includes periorbital region samples and additional site samples from the same sample 3D model. The periorbital region base is reconstructed into a periorbital region model according to the target fitting coefficient, and the additional region base is reconstructed into an additional region model using the target fitting coefficient; The periorbital region model and the additional region model are fused with the 3D face model to be processed, and the fused edges are smoothed to obtain the target 3D face model.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the steps of the face 3D model processing method according to any one of claims 1 to 5 through the computer program.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the face 3D model processing method as described in any one of claims 1 to 5.

Images