CN111753652B - Three-dimensional face recognition method based on data enhancement - Google Patents
Three-dimensional face recognition method based on data enhancement Download PDFInfo
- Publication number
- CN111753652B CN111753652B CN202010406689.3A CN202010406689A CN111753652B CN 111753652 B CN111753652 B CN 111753652B CN 202010406689 A CN202010406689 A CN 202010406689A CN 111753652 B CN111753652 B CN 111753652B
- Authority
- CN
- China
- Prior art keywords
- face
- data
- dimensional
- dimensional face
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000012549 training Methods 0.000 claims abstract description 19
- 238000012360 testing method Methods 0.000 claims abstract description 12
- 238000007781 pre-processing Methods 0.000 claims abstract description 10
- 238000013461 design Methods 0.000 claims abstract description 5
- 238000013528 artificial neural network Methods 0.000 claims abstract 2
- 238000013527 convolutional neural network Methods 0.000 claims description 29
- 238000012545 processing Methods 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 7
- 238000001914 filtration Methods 0.000 claims description 6
- 238000003062 neural network model Methods 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 230000001965 increasing effect Effects 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 238000003672 processing method Methods 0.000 claims description 2
- 238000012216 screening Methods 0.000 claims description 2
- 238000005457 optimization Methods 0.000 abstract description 2
- 238000005516 engineering process Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 4
- 230000006978 adaptation Effects 0.000 description 2
- 230000001815 facial effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
- G06V40/165—Detection; Localisation; Normalisation using facial parts and geometric relationships
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Abstract
The invention discloses a three-dimensional face recognition method based on data enhancement, which comprises the following steps: s1: data enhancement; s2: preprocessing data; s3: carrying out parameter and network structure optimization design by using a VGG-16 neural network basic model, and constructing a special recognition network for the three-dimensional face; s4: dividing a data set into a training set and a testing set, training a network model by using the training set, and testing the performance of the recognition algorithm on the testing set; s5: removing the last full link layer by using the trained recognition network model to obtain a face feature vector, and storing the face feature vector for identity judgment; and S6, acquiring the feature vector of the registered face, comparing the feature vector with the features stored in the data set, and identifying the identity of the target person. The invention can automatically learn to obtain the effective characteristics of the three-dimensional face, and particularly can solve the learning problems of overfitting of an identification network, poor applicability and the like caused by difficulty in obtaining the three-dimensional data of the real face in the three-dimensional face identification.
Description
Technical Field
The invention relates to the technical fields of computer vision, face recognition, machine learning and the like, in particular to a three-dimensional face recognition method based on data enhancement.
Background
Face recognition is a highly non-invasive biometric technology that is rapidly becoming the first tool in areas of surveillance (e.g., border control, suspect tracking, identification), security (e.g., system entry, banking, file encryption), entertainment (e.g., human-machine interaction, three-dimensional animation, virtual reality), and the like. Advances in deep learning have led to revolutionary improvements in various computer vision tasks, where CNN-based face recognition is considered to have surpassed human performance.
Although two-dimensional face recognition has excellent performance and data usability, two-dimensional face recognition is still challenged by illumination, pose, and scale changes. Furthermore, facial texture is not always stable to identity, as it varies with makeup and age. Three-dimensional face recognition techniques have the ability to address these problems, but no commonly recognized dedicated convolutional neural network model for three-dimensional face data has been disclosed in prior studies. The main reason is that a large amount of three-dimensional face training and testing data is lacked, the three-dimensional face data cannot be automatically obtained through a web crawler like two-dimensional face data, and the acquisition and construction of a three-dimensional large-scale data set also need to spend huge time and labor cost. It is currently known that the largest publicly available three-dimensional face data set ND-2006 (superset of FRGCv 2) only scans 888 unique identities and takes more than two years to collect and build, while the publicly largest two-dimensional face data set contains 1000 million face pictures of nearly 10 million people.
The data enhancement aims to solve the problem that labeled three-dimensional face data used for training a convolutional neural network are insufficient. Creating a composite face from an existing three-dimensional face model is limited to the linear space of the particular model, resulting in limited shape variations of the face. And by introducing the expression change method, more scans can be generated for each object without increasing the number of unique identities in the data.
In recent years, face reconstruction technology has become mature, the iteration of the early method based on 3DMM model reconstruction coefficient takes a lot of time, and the model-based method makes the reconstructed face limited in geometry. Therefore, researchers design a new UV position map to represent the face, and learn corresponding characteristics by using a convolutional neural network, so that an end-to-end face reconstruction process based on a single picture is realized. The method can achieve the purpose of real-time conversion, and can obtain complete three-dimensional face information.
Therefore, by fully utilizing the advantages of the method, the human face reconstruction technology can be used for enhancing the three-dimensional human face data set, a large amount of three-dimensional human face data with the same identity as the two-dimensional data are obtained, and the existing small-scale three-dimensional human face database is expanded. Subsequently, the enhanced data set can be used for training a convolutional neural network to acquire the characteristics of the three-dimensional face, so that an efficient and reliable three-dimensional face recognition system is realized.
Disclosure of Invention
Based on the above, the present invention is directed to a three-dimensional face recognition method based on data enhancement, which can directly obtain three-dimensional feature information of a target face by using three-dimensional point cloud data of a single face as input data, so as to perform identity comparison. The method mainly solves the problems of overfitting, poor recognition performance and the like easily occurring in the process of training the convolutional neural network by using a small-scale database, so that a three-dimensional face recognition model which can be applied in a large-scale data scene is obtained, and the requirement of recognition accuracy on a face recognition algorithm is better met.
In order to achieve the above object, the present invention provides a three-dimensional face recognition method based on data enhancement, which includes the following steps:
s1, data enhancement, namely reconstructing a three-dimensional face based on an easily-obtained 2D face picture, so that the number of identities in an original three-dimensional data set is increased, and the enhancement and expansion of three-dimensional face data are realized;
s2, data preprocessing, namely acquiring point cloud data in a face area by using a face preprocessing method, constructing three-channel data, carrying out normalization processing, mapping the three-channel data to RGB three-channel data, and inputting the three-channel data serving as data of a convolutional neural network in a three-channel mode;
s3, designing and optimizing internal parameters and a network structure based on a VGG-16 neural network model, adjusting the size of a convolution kernel, and constructing a special deep learning identification network model aiming at three-dimensional face identification;
s4, training a model, namely dividing a training set and a test set, training the convolutional neural network by taking data as input, and verifying the performance of the recognition algorithm on the test set;
s5, removing the last full-link layer based on the trained convolutional neural network, taking the full-link layer as a feature extractor, extracting the human face features, and storing the features;
and S6, registering a new face, acquiring a feature vector, comparing the feature vector with the features stored in the data set, and identifying the identity.
Further, in step S1, the reconstructing a three-dimensional face based on an easily obtained 2D face picture means that, for each 2D face picture, a three-dimensional face point cloud data corresponding to an identity is reconstructed by using a three-dimensional face reconstruction technique.
Further, in step S2, the obtaining of the point cloud data in the face region by using the face preprocessing method means that the obtained three-dimensional face point cloud data is preprocessed, after the nose tip position is found by a geometric method, the face region range is determined by a fixed radius, the face region range is screened out, and interpolation and filtering are performed on the face region range to obtain the point cloud data of the face part.
Further, in step S2, the constructing three-channel data, performing normalization processing, mapping to RGB three-channel data, and using a three-channel mode as data input of a convolutional neural network means that (x, y, z) of the three-dimensional point cloud data of the human face is converted, and a depth value, an elevation angle and an azimuth angle (D, a, E) are calculated, wherein the depth is a distance from a pixel point to a viewpoint, the elevation angle is an included angle of an XY plane corresponding to a line connecting the nasal cusps, and the azimuth angle starts to be calculated clockwise in the positive direction of the y axis; and after conversion, carrying out normalization processing on the data of the three channels, corresponding to the RGB space, and inputting the data serving as the data of the convolutional neural network.
Further, in step S3, the design and optimization of the internal parameters and the network structure are performed based on the VGG-16 neural network model, and the adjustment of the size of the convolution kernel means that the network parameters are randomly initialized based on the VGG-16 network structure, and the size of the convolution kernel is optimized from 3x3 to 7x7, so that the three-dimensional point cloud data is fully utilized, and the local structural features of the three-dimensional face point cloud are better learned.
Further, step S6 is specifically to, when new face data comes, extract features using the network trained in step S5, compare the extracted features with face features already stored in the database through mean square error calculation, find the minimum mean square error, that is, the corresponding identity, and complete the recognition work.
Further, in step S2, the face processing method includes nose tip detection, face segmentation, interpolation and filtering.
The method provided by the invention has important significance for providing a novel data enhancement mode and improving the accuracy of three-dimensional face recognition, and the beneficial effects are embodied in the following aspects:
1. the invention applies the convolutional neural network to the three-dimensional face recognition, can automatically learn to obtain the most effective characteristics for recognition, and does not need manual participation.
2. The invention is a method based on a convolutional neural network, omits the complicated manual design flow of the traditional face recognition, can directly recognize the corresponding identity information by utilizing the input three-dimensional face point cloud data, and automatically completes the face recognition.
3. In order to enable the convolutional neural network to achieve higher identification accuracy rate, a large amount of labeled data is needed, the invention converts the 2D face image which is easy to obtain into three-dimensional face data through face reconstruction, effectively enlarges the sample space, and can solve the over-fitting problem which is easy to occur when a small-scale database trains the convolutional neural network.
Due to the advantages, the method can identify the point cloud data of the three-dimensional face, effectively improve the accuracy and the robustness of a three-dimensional face identification system, and can be applied to an entrance guard and security system with a three-dimensional face acquisition camera and a mobile device with a three-dimensional camera.
Drawings
FIG. 1 is a flow chart of a three-dimensional face recognition method based on data enhancement according to the present invention;
fig. 2 is a three-dimensional face reconstruction result based on a single face image;
fig. 3 shows the result of preprocessing and converting three-dimensional face data into corresponding RGB three-channel data.
Detailed Description
It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.
The invention is described in further detail below with reference to the figures and the specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
In practical applications, although 2D face recognition has excellent performance and data usability, 2D face recognition is still challenged by changes in lighting, pose, and scale. Furthermore, facial texture is not always stable to identity, as it changes with makeup. Three-dimensional face recognition is likely to address these shortcomings,
however, the convolutional neural network-based method lacks a large number of data sets as support, so that the network is over-fitted and cannot achieve a high recognition rate in an open environment. Most of the data enhancement methods proposed by the predecessors are changes on the original identity, and a simple, convenient and effective acquisition mode of the face information of the new identity is lacked. Therefore, the invention provides a three-dimensional face recognition method based on data enhancement, which realizes data enhancement on the basis of a three-dimensional face reconstruction technology of a single face image, increases new identities, constructs a simple and efficient three-dimensional face recognition system, and solves the problem of overfitting easily caused by training a convolutional neural network by a small-scale data set while improving the accuracy.
The invention provides a three-dimensional face recognition method based on data enhancement, the flow chart of which is shown in figure 1, and the method comprises the following steps:
and S1, data enhancement, namely reconstructing a three-dimensional face based on a 2D face picture which is easy to obtain, so that the face identity in a data set is increased, and the data enhancement effect is realized.
The step S1 specifically includes: and performing three-dimensional face reconstruction by using the 2D face data set, and reconstructing a complete three-dimensional face point cloud data corresponding to the identity by using a three-dimensional face reconstruction technology aiming at an independent 2D face picture, wherein the obtained point cloud comprises complete face geometric information.
And S2, preprocessing three-dimensional face data, including methods of nose tip detection, face segmentation, interpolation, filtering and the like, to obtain point cloud data in a face range.
The step S2 specifically comprises the following steps: the method comprises the steps of preprocessing obtained three-dimensional face point cloud data, firstly, making tangent planes on the point cloud data, drawing circles aiming at each point on the tangent planes, intersecting the circles with the other two points to form triangles, calculating the heights of the triangles, reserving 5 highest points corresponding to the triangles aiming at each tangent plane, comprehensively comparing and selecting the highest points as confidence points of nose tips, then determining the range of a face area by using a fixed radius, screening out point clouds in the range, and performing interpolation and filtering processing on the point clouds to obtain the point cloud data of a simple face part. And constructing three-channel data, carrying out normalization processing, mapping the three-channel data to RGB three-channel data, and inputting the three-channel data into the convolutional neural network in a three-channel mode. The method specifically comprises the following steps: and (x, y, z) of the human face three-dimensional point cloud data is converted, and the depth value, the elevation angle and the azimuth angle (D, A, E) are obtained through calculation, wherein the depth is the distance from a pixel point to a viewpoint, the elevation angle is an included angle of an XY plane corresponding to a line connecting the nose tip point, and the azimuth angle is calculated clockwise in the positive direction of the y axis. And after conversion, carrying out normalization processing on the data of the three channels, and corresponding to the RGB space. The resulting data size is 224x224x3 as input data to the convolutional neural network.
And S3, optimizing model parameters and a network structure based on the VGG-16 neural network model, adjusting a convolution kernel, and constructing a special recognition network for the three-dimensional face.
The step S3 specifically comprises the following steps: based on the VGG-16 network structure, network parameters are initialized randomly, and the size of a convolution kernel is replaced from the original 3x3 to 7x7, so that the local information of the three-dimensional point cloud data is fully utilized, and the local structure characteristics of the three-dimensional point cloud are better learned.
S4, training a model, namely dividing a training set and a test set, training the convolutional neural network by taking data as input, and verifying the recognition rate on the test set;
the step S4 specifically comprises the following steps: the data set is as follows 4:1, the convolutional neural network is divided into a training set and a test set, the proper learning rate and the parameter of the batch size training convolutional neural network are adjusted, and the verification is carried out on the test set.
And S5, removing the last full-link layer based on the trained convolutional neural network, and taking the full-link layer as a feature extractor to extract the human face features.
The step S5 specifically comprises the following steps: and (3) reserving the convolution layer of the trained convolutional neural network model, removing the last full-link layer, and using the convolution layer as a feature extractor for obtaining and storing the identity features of the face data set.
And S6, registering a new face, acquiring a feature vector, comparing the feature vector with the features stored in the data set, and identifying the identity.
The step S6 specifically includes: when new face data comes, the network is used for extracting features, the features are compared with the mean square error of face features stored in a database, the minimum mean square error is found, namely the corresponding identity is found, and the recognition work is finished.
Examples of the invention are listed below:
application example 1: the data enhancement-based three-dimensional face recognition method is applied to a security system.
The invention can be widely applied to security scenes for identity authentication and identification by using human faces. For example, a theft case occurs at night in a place, the criminal process of a criminal is recorded and shot by a monitoring system, and the identity of the criminal cannot be effectively confirmed by using 2D face recognition due to the fact that the light of a crime scene is dim. The police installs and configures a three-dimensional camera in advance for monitoring through the three-dimensional face recognition method. And then, the identity of the criminal suspect can be immediately confirmed by extracting the characteristics of the three-dimensional face information and comparing the characteristics with the face information in the police database. Due to the assistance of the security system based on the three-dimensional face recognition, the police can lock the criminal suspect within hours and can cover more scenes in which the 2D face recognition cannot be applied.
Application example 2: the three-dimensional face recognition method based on data enhancement is applied to the mobile terminal.
The invention can be applied to mobile terminals. Mobile devices are widely used in daily life of people, typical applications include mobile phone payment, storing private information, and the like, and how to ensure the security of mobile devices is receiving more and more attention. With the development of the biometric technology, compared with the traditional methods such as password input, the method has the advantages of good user friendliness, high reliability and the like, and the human face becomes a mainstream technology for guaranteeing the safety of the mobile device due to the characteristic that the human face does not need to be contacted. However, the quality of the face image acquired by the mobile terminal is low, and factors such as high noise, low resolution, defocusing, motion blur and the like exist, and the face image is seriously influenced by light rays and often cannot be identified. In addition, 2D face recognition on mobile devices is vulnerable to false faces. The invention utilizes the face reconstruction to enhance the data set, improves the identification accuracy and enhances the robustness, and the invention is an identification method based on the convolutional neural network, and omits the complex process of characteristic extraction and classification in the traditional method, so the invention has higher efficiency and is especially suitable for the application of mobile terminals with less registration and identification samples.
The technical means not described in detail in the present application are known techniques.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, many modifications and adaptations can be made without departing from the principle of the present invention, and such modifications and adaptations should also be considered as the scope of the present invention.
Claims (5)
1. A three-dimensional face recognition method based on data enhancement is characterized by comprising the following steps:
s1, data enhancement, namely reconstructing a three-dimensional face based on an easily-obtained 2D face picture, so that the number of identities in an original three-dimensional data set is increased, and the enhancement and expansion of three-dimensional face data are realized;
s2, data preprocessing, namely acquiring point cloud data in a face area by using a face preprocessing method, constructing three-channel data, carrying out normalization processing, mapping the three-channel data to RGB three-channel data, and inputting the three-channel data serving as data of a convolutional neural network in a three-channel mode;
s3, designing and optimizing internal parameters and a network structure based on a VGG-16 neural network model, adjusting the size of a convolution kernel, and constructing a special deep learning identification network model aiming at three-dimensional face identification;
s4, training a model, namely dividing a training set and a test set, training the convolutional neural network by taking data as input, and verifying the performance of the recognition algorithm on the test set;
s5, removing the last full-link layer based on the trained convolutional neural network, taking the full-link layer as a feature extractor, extracting the human face features, and storing the human face features;
s6, registering a new face, acquiring a feature vector, comparing the feature vector with features stored in a data set, and identifying an identity;
wherein, the step S2 specifically comprises: preprocessing the obtained three-dimensional face point cloud data, firstly, making tangent planes on the point cloud data, drawing circles aiming at each point on the tangent planes, intersecting the circles with the other two points to form triangles and calculating the heights of the triangles, reserving 5 highest points corresponding to the triangles aiming at each tangent plane, comprehensively comparing and selecting the highest points as confidence points of nose tips, then determining the range of a face area by using a fixed radius, screening out the point clouds in the range, and performing interpolation and filtering processing on the point clouds to obtain the point cloud data of a simple face part; converting (x, y, z) of the human face three-dimensional point cloud data, and calculating to obtain a depth value, an elevation angle and an azimuth angle (D, A, E), wherein the depth is the distance from a pixel point to a viewpoint, the elevation angle is an included angle of an XY plane corresponding to a line connecting with a nose tip point, and the azimuth angle is calculated clockwise in the positive direction of a y axis; and after conversion, carrying out normalization processing on the data of the three channels, corresponding to the RGB space, and taking the finally obtained data as input data of the convolutional neural network.
2. The method for three-dimensional face recognition based on data enhancement as claimed in claim 1, wherein in step S1, the step of reconstructing the three-dimensional face based on the easily obtained 2D face pictures is to reconstruct a three-dimensional face point cloud data of the corresponding identity for each 2D face picture by using a three-dimensional face reconstruction technique.
3. The method for identifying three-dimensional human faces based on data enhancement according to claim 1, wherein in step S3, the VGG-16 neural network model is used to design and optimize internal parameters and network structures, and the adjusting of the size of the convolution kernel means that the network parameters are randomly initialized and the size of the convolution kernel is optimized from 3x3 to 7x7 based on the VGG-16 neural network structure, so as to make full use of the three-dimensional point cloud data and better learn the local structural features of the three-dimensional point cloud of human faces.
4. The data enhancement-based three-dimensional face recognition method according to claim 1, wherein the step S6 is specifically that when new face data comes, the network trained in the step S5 is used to extract features, and the features are compared with the face features already stored in the database through mean square error calculation, so that the minimum mean square error is found, that is, the corresponding identity is found, and the recognition work is completed.
5. The method for recognizing three-dimensional human faces based on data enhancement according to claim 1, wherein in the step S2, the human face processing method comprises nose tip detection, human face segmentation and interpolation and filtering.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010406689.3A CN111753652B (en) | 2020-05-14 | 2020-05-14 | Three-dimensional face recognition method based on data enhancement |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010406689.3A CN111753652B (en) | 2020-05-14 | 2020-05-14 | Three-dimensional face recognition method based on data enhancement |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111753652A CN111753652A (en) | 2020-10-09 |
| CN111753652B true CN111753652B (en) | 2022-11-29 |
Family
ID=72673861
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010406689.3A Active CN111753652B (en) | 2020-05-14 | 2020-05-14 | Three-dimensional face recognition method based on data enhancement |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111753652B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113807182B (en) * | 2021-08-17 | 2024-02-27 | 北京地平线信息技术有限公司 | Method, device, medium and electronic equipment for processing point cloud |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104504410A (en) * | 2015-01-07 | 2015-04-08 | 深圳市唯特视科技有限公司 | Three-dimensional face recognition device and method based on three-dimensional point cloud |
| CN108537191A (en) * | 2018-04-17 | 2018-09-14 | 广州云从信息科技有限公司 | A kind of three-dimensional face identification method based on structure light video camera head |
| CN108805056A (en) * | 2018-05-29 | 2018-11-13 | 电子科技大学 | A kind of monitoring camera-shooting face sample extending method based on 3D faceforms |
| CN109255340A (en) * | 2018-10-29 | 2019-01-22 | 东北大学 | It is a kind of to merge a variety of face identification methods for improving VGG network |
| CN109753875A (en) * | 2018-11-28 | 2019-05-14 | 北京的卢深视科技有限公司 | Face identification method, device and electronic equipment based on face character perception loss |
| CN110147721A (en) * | 2019-04-11 | 2019-08-20 | 阿里巴巴集团控股有限公司 | A kind of three-dimensional face identification method, model training method and device |
| CN110188621A (en) * | 2019-05-09 | 2019-08-30 | 东南大学 | A kind of three-dimensional face expression recognition methods based on SSF-IL-CNN |
| CN110781924A (en) * | 2019-09-29 | 2020-02-11 | 哈尔滨工程大学 | Side-scan sonar image feature extraction method based on full convolution neural network |
-
2020
- 2020-05-14 CN CN202010406689.3A patent/CN111753652B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104504410A (en) * | 2015-01-07 | 2015-04-08 | 深圳市唯特视科技有限公司 | Three-dimensional face recognition device and method based on three-dimensional point cloud |
| CN108537191A (en) * | 2018-04-17 | 2018-09-14 | 广州云从信息科技有限公司 | A kind of three-dimensional face identification method based on structure light video camera head |
| CN108805056A (en) * | 2018-05-29 | 2018-11-13 | 电子科技大学 | A kind of monitoring camera-shooting face sample extending method based on 3D faceforms |
| CN109255340A (en) * | 2018-10-29 | 2019-01-22 | 东北大学 | It is a kind of to merge a variety of face identification methods for improving VGG network |
| CN109753875A (en) * | 2018-11-28 | 2019-05-14 | 北京的卢深视科技有限公司 | Face identification method, device and electronic equipment based on face character perception loss |
| CN110147721A (en) * | 2019-04-11 | 2019-08-20 | 阿里巴巴集团控股有限公司 | A kind of three-dimensional face identification method, model training method and device |
| CN110188621A (en) * | 2019-05-09 | 2019-08-30 | 东南大学 | A kind of three-dimensional face expression recognition methods based on SSF-IL-CNN |
| CN110781924A (en) * | 2019-09-29 | 2020-02-11 | 哈尔滨工程大学 | Side-scan sonar image feature extraction method based on full convolution neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111753652A (en) | 2020-10-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106780906B (en) | A kind of testimony of a witness unification recognition methods and system based on depth convolutional neural networks | |
| CN108520216B (en) | Gait image-based identity recognition method | |
| US9117105B2 (en) | 3D face recognition method based on intermediate frequency information in geometric image | |
| CN110348330B (en) | Face pose virtual view generation method based on VAE-ACGAN | |
| CN108537191B (en) | Three-dimensional face recognition method based on structured light camera | |
| Dagnes et al. | Occlusion detection and restoration techniques for 3D face recognition: a literature review | |
| Feng et al. | Robust and efficient algorithms for separating latent overlapped fingerprints | |
| CN109145745B (en) | Face recognition method under shielding condition | |
| CN107145842A (en) | With reference to LBP characteristic patterns and the face identification method of convolutional neural networks | |
| CN107392187B (en) | Face in-vivo detection method based on gradient direction histogram | |
| Medioni et al. | Identifying noncooperative subjects at a distance using face images and inferred three-dimensional face models | |
| CN108681737B (en) | Method for extracting image features under complex illumination | |
| Kim et al. | A survey and proposed framework on the soft biometrics technique for human identification in intelligent video surveillance system | |
| CN108573231B (en) | Human body behavior identification method of depth motion map generated based on motion history point cloud | |
| CN105956515A (en) | Stereo-hyperspectral human face recognition method based on auroral imaging | |
| AU2012211974A1 (en) | Systems and methods for biometric identification | |
| CN107169479A (en) | Intelligent mobile equipment sensitive data means of defence based on fingerprint authentication | |
| CN101464945A (en) | Identification characteristic identification method based on finger back arthrosis veins | |
| CN111126240A (en) | Three-channel feature fusion face recognition method | |
| Zhao et al. | Trustworthy authorization method for security in Industrial Internet of Things | |
| CN114218543A (en) | Encryption and unlocking system and method based on multi-scene expression recognition | |
| CN111753652B (en) | Three-dimensional face recognition method based on data enhancement | |
| CN112686202B (en) | Human head identification method and system based on 3D reconstruction | |
| Zhao et al. | The application and implementation of face recognition in authentication system for distance education | |
| CN112380966B (en) | Monocular iris matching method based on feature point re-projection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP01 | Change in the name or title of a patent holder | ||
| CP01 | Change in the name or title of a patent holder |
Address after: 300072 Tianjin City, Nankai District Wei Jin Road No. 92 Patentee after: Tianjin University Patentee after: Tianjin Zhongke intelligent identification Co.,Ltd. Address before: 300072 Tianjin City, Nankai District Wei Jin Road No. 92 Patentee before: Tianjin University Patentee before: TIANJIN ZHONGKE INTELLIGENT IDENTIFICATION INDUSTRY TECHNOLOGY RESEARCH INSTITUTE Co.,Ltd. |