CN111339833B - Identity verification method, system and equipment based on face edge calculation - Google Patents
Identity verification method, system and equipment based on face edge calculation Download PDFInfo
- Publication number
- CN111339833B CN111339833B CN202010078951.6A CN202010078951A CN111339833B CN 111339833 B CN111339833 B CN 111339833B CN 202010078951 A CN202010078951 A CN 202010078951A CN 111339833 B CN111339833 B CN 111339833B
- Authority
- CN
- China
- Prior art keywords
- face
- layer
- weight
- training
- 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
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
- G06V40/172—Classification, e.g. identification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
-
- 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
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/063—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using electronic means
-
- 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- G06V40/168—Feature extraction; Face representation
Abstract
The invention provides an identity verification method based on face edge calculation, which comprises the following steps: s1, installing identity verification equipment on terminal equipment as an edge node of a network; s2, uploading a certain number of face training samples to a cloud server, and training a convolutional neural network of the cloud server by using the face training samples; s3, integrating weights of all layers of a convolutional neural network of the cloud server to obtain a weight configuration script, and issuing the weight configuration script to an edge node through the network; and S4, downloading a weight configuration script by the identity verification equipment, and adjusting the weight of each layer by using a face recognizer of a convolutional neural network in the identity verification equipment according to the weight configuration script. The requirement on the terminal equipment is reduced, meanwhile, the increase of communication cost and delay time in the recognition process is avoided, in addition, a secondary training adjustment process for the weight is added, and the accuracy and the adaptability of the face recognizer in face recognition are improved.
Description
Technical Field
The invention relates to the technical field of convolutional neural networks, in particular to an identity verification method, system and device based on face edge calculation.
Background
At present, the application scenarios of the identity verification technology based on face recognition are more and more abundant, such as face access control, face door lock, newspapers and periodicals or express delivery locker based on face recognition in the community, household appliances and smart homes which realize personalized control through face recognition in the family, and the like; specifically, as for face recognition, there are many classical algorithms, and a convolutional neural network algorithm has a higher accuracy and a stronger robustness, and therefore, attention is paid more and more, a face recognizer of a convolutional neural network includes two processes of training and recognition, wherein the training process includes forward propagation of information and backward propagation of errors, that is, a certain number of face samples are used as input training data, forward propagation is performed layer by layer in the neural network to obtain an output error of the face recognizer, and whether the error is within an acceptable interval is judged, if the error is not within the interval, the backward propagation of the error is performed, weights are adjusted layer by layer, and iteration is performed repeatedly until the error reaches the acceptable interval, so that training of the face recognizer of the convolutional neural network is completed, and the face recognizer of the convolutional neural network after training can perform person identity verification according to the input face data.
However, the training of the convolutional neural network needs a large data processing amount and a strong hardware performance, the terminal devices in the existing smart community and smart home cannot meet the requirements, if only the terminal devices are used as devices for front-end acquisition, all data are transmitted to the server at the cloud end to perform training and recognition of the convolutional neural network, the communication cost is increased, and the delay time is prolonged.
Therefore, how to compromise and adjust the terminal device and the cloud server, reduce the requirements on the terminal device, and avoid the increase of communication cost and delay time in the identification process is a problem to be solved urgently by those skilled in the art.
Disclosure of Invention
In view of the above, the present invention provides an identity verification method, system and device based on face edge calculation, which utilize a cloud server to train a convolutional neural network, extract weights of each layer of the convolutional neural network after training, send the weights to an identity verification device installed in a terminal device through a network, and further adjust weights of each layer of a face recognizer which is built in the identity verification device and adopts the convolutional neural network, so as to separate the recognition and training processes of the convolutional neural network, reduce requirements on the terminal device, and avoid the increase of communication cost and delay time of the recognition process.
In order to achieve the purpose, the invention adopts the following technical scheme:
an identity verification method based on face edge calculation comprises the following steps:
s1, installing identity verification equipment on terminal equipment as an edge node of a network;
s2, uploading a certain number of face training samples to a cloud server, training the convolutional neural network of the cloud server by using the face training samples, and acquiring the trained convolutional neural network of the cloud server;
s3, integrating weights of all layers of a convolutional neural network of the cloud server to obtain a weight configuration script, and issuing the weight configuration script to an edge node through the network;
and S4, downloading the weight configuration script by the identity verification equipment, and adjusting the weights of all layers by adopting a face recognizer of a convolutional neural network in the identity verification equipment according to the weight configuration script.
Specifically, the convolutional neural network comprises an input layer, a convolutional layer, a pooling layer, a full-link layer and a classification output layer, the face pictures are transmitted layer by using the convolutional neural network and processed, and the obtained face recognition result is high in accuracy and strong in robustness.
Specifically, the specific process of training the neural network in S2 is as follows: acquiring a face training sample from the input layer and transmitting the face training sample to the convolutional layer; extracting the characteristics of the face training sample by the convolutional layer and transmitting the characteristics to the pooling layer; the pooling layer performs pooling operation on the extracted features of the face training samples, reduces the dimensionality of feature data, and transmits the extracted features to the full-connection layer; the full-connection layer classifies the extracted features of the face training sample to obtain a classification model, and transmits the classification model to the classification output layer to output a face recognition result; comparing the output face recognition result with the expected value to obtain the error between the output result and the expected value; and reversely transmitting the error, calculating the error of each layer, updating the weight of each layer according to the error until the output face recognition result is in an interval acceptable by an expected value, and obtaining a trained convolutional neural network capable of carrying out face recognition.
Preferably, the method further comprises the following steps: the identity verification equipment carries out autonomous training on the face recognizer according to a face image acquired on site to obtain a weight value adjusting coefficient; and on the basis of the weight obtained by S4 adjustment, multiplication operation is carried out on the weight and the adjustment coefficient, and secondary training adjustment of the weight is completed. After the time passes, the face changes to a certain extent, so that the weight obtained through S4 adjustment is subjected to secondary training adjustment, namely the weight is updated in real time according to the face collected on site, and the accuracy and the adaptability of the face recognizer are improved.
Based on the method, the following system is designed:
an identity verification system based on face edge calculation comprises a configuration module, identity verification equipment and a cloud server; wherein the content of the first and second substances,
the identity verification equipment comprises a weight value adjusting unit, a weight value configuration unit and a face recognizer;
the cloud server comprises a virtual training unit and an integration unit;
the configuration module is used for installing the identity verification equipment on the terminal equipment as an edge node of a network;
the virtual training unit is used for acquiring a certain number of face training samples, training the convolutional neural network of the cloud server by using the face training samples, and acquiring the trained convolutional neural network of the cloud server;
the integration unit is used for integrating weights of all layers of the convolutional neural network of the cloud server to obtain a weight configuration script and sending the weight configuration script to the edge node through the network;
the weight configuration unit is used for downloading a weight configuration script, and the weight adjustment unit is used for adjusting the weights of all layers of the face recognizer which adopts a convolutional neural network in the identity verification equipment according to the weight configuration script.
Specifically, the convolutional neural network comprises an input layer, a convolutional layer, a pooling layer, a full-link layer and a classification output layer.
Specifically, the specific process of training the neural network by the virtual training unit is as follows: acquiring a face training sample from the input layer and transmitting the face training sample to the convolutional layer; the convolutional layer extracts the characteristics of the face training sample and transmits the characteristics to the pooling layer; the pooling layer performs pooling operation on the extracted features of the face training samples, reduces the dimensionality of feature data, and transmits the extracted features to the full-connection layer; the full-connection layer classifies the extracted features of the face training samples to obtain a classification model, transmits the classification model to the classification output layer and outputs a face recognition result; comparing the output face recognition result with the expected value to obtain the error between the output result and the expected value; and reversely transmitting the error, calculating the error of each layer, updating the weight of each layer according to the error until the output face recognition result is in an interval acceptable by an expected value, and obtaining a trained convolutional neural network capable of carrying out face recognition.
Preferably, the identity verification device further comprises a secondary training adjusting unit, wherein the secondary training adjusting unit is used for performing autonomous training on the face recognizer according to a face image acquired on site to obtain an adjusting coefficient of the weight; and on the basis of the weight value obtained by the weight value adjusting unit, multiplying the weight value by an adjusting coefficient to finish secondary training and adjusting of the weight value.
An identity verification device based on face edge calculation comprises a weight value configuration unit, a weight value adjusting unit and a face recognizer; wherein the content of the first and second substances,
the weight configuration unit is used for downloading and storing a weight configuration script from a cloud end;
and the weight value adjusting unit is used for adjusting the weight values of all layers of the face recognizer which adopts a convolutional neural network in the identity verification equipment by referring to the weight value configuration script.
Preferably, the system further comprises a secondary training adjusting unit, wherein the secondary training adjusting unit is used for performing autonomous training on the face recognizer according to a face image acquired on site to obtain an adjusting coefficient of the weight; and on the basis of the weight value obtained by the weight value adjusting unit, multiplying the weight value by an adjusting coefficient to finish secondary training and adjusting of the weight value.
The invention has the following beneficial effects:
based on the technical scheme, the invention provides the identity verification method, the identity verification system and the identity verification equipment based on face edge calculation, which are based on the prior art, the requirements on terminal equipment are reduced by separating the identification and training processes of a convolutional neural network, meanwhile, the increase of communication cost and delay time of the identification process is avoided, in addition, the secondary training adjustment process of a weight is added, and the accuracy and the adaptability of a face recognizer in face recognition are improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
FIG. 1 is a flow chart of a method for identity verification based on face edge calculation;
FIG. 2 is a block diagram of a system for identity verification based on face edge calculation;
fig. 3 is a schematic structural diagram of an identity verification device based on face edge calculation.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, the present invention provides the following method:
an identity verification method based on face edge calculation comprises the following steps:
s1, installing identity verification equipment on terminal equipment as an edge node of a network;
specifically, the identity verification equipment is installed on terminal equipment of a smart community and a smart home, is connected with the cloud server through a network, and is used for receiving a weight configuration script of a convolutional neural network obtained by training a face training sample by the cloud server.
S2, uploading a certain number of face training samples to a cloud server, training the convolutional neural network of the cloud server by using the face training samples, and acquiring the trained convolutional neural network of the cloud server;
specifically, a user side is further arranged, a certain number of face training samples can be uploaded to the cloud server through the user side by the user side, for example, if the user wishes to train a face door lock of the user, a certain number of photos and videos of each member of the user can be used as the face training samples, the face training samples are uploaded to the cloud server through a human-computer interaction system of the user side through a network, and the face training samples of each member are used for training the convolutional neural network one by one; in addition, the human-computer interaction system at the user side can also guide uploading of the photos and videos of the face training samples to a certain extent, for example, guide the user to upload the photos and videos of different face angles.
Specifically, the convolutional neural network includes input layer, convolution layer, pooling layer, full tie layer and classification output layer, and the process that face training sample trained at the high in the clouds server is: acquiring a face training sample from the input layer and transmitting the face training sample to the convolutional layer; the convolutional layer extracts the characteristics of the face training sample and transmits the characteristics to the pooling layer; the pooling layer performs pooling operation on the extracted features of the face training samples, reduces the dimensionality of feature data, and transmits the extracted features to the full-connection layer; the full-connection layer classifies the extracted features of the face training samples to obtain a classification model, transmits the classification model to the classification output layer and outputs a face recognition result; comparing the output face recognition result with the expected value to obtain the error between the output result and the expected value; and reversely transmitting the error, calculating the error of each layer, updating the weight of each layer according to the error until the output face recognition result is in an interval acceptable by an expected value, and obtaining a trained convolutional neural network capable of carrying out face recognition.
S3, integrating weights of all layers of a convolutional neural network of the cloud server to obtain a weight configuration script, and issuing the weight configuration script to an edge node through the network;
and the cloud server extracts weights of all layers of the convolutional neural network obtained by training and integrates the weights into a weight configuration script.
And S4, downloading a weight configuration script by the identity verification equipment, and adjusting the weight of each layer by using a face recognizer of a convolutional neural network in the identity verification equipment according to the weight configuration script.
Specifically, the identity verification equipment comprises a weight adjusting unit and a weight configuration unit, wherein the weight configuration unit receives and downloads a weight configuration script integrated by a cloud server through a network, the weight adjusting unit obtains weights of all layers of a convolutional neural network in the weight configuration script from the weight configuration unit, adjusts the weight of a built-in face recognizer adopting the convolutional neural network of the identity verification equipment according to the weights, and obtains a face recognizer capable of carrying out face recognition.
In order to further optimize the technical characteristics, the method further comprises the following steps: the identity verification equipment carries out autonomous training on the face recognizer according to a face image acquired on site, namely, a face picture acquired on site is propagated in a convolutional neural network of the face recognizer in a forward direction, a face recognition result is output after processing and feature extraction, the face picture is compared with an expected result to obtain an adjusting coefficient of a weight, the adjusting coefficient is propagated in a backward direction and is multiplied by an original weight, so that the adjustment of weights of all layers of the convolutional neural network is realized according to the face picture acquired on site, the face recognizer is adjusted in time according to the real-time change of the face, and the secondary training adjustment of the weight is completed.
Based on the method, the following system is set:
as shown in fig. 2, an identity verification system based on face edge calculation includes a configuration module 1, an identity verification device 2, and a cloud server 3; wherein the content of the first and second substances,
the identity verification device 2 comprises a weight value adjusting unit 22, a weight value configuration unit 21 and a face recognizer 23;
the cloud server 3 comprises a virtual training unit 31 and an integration unit 32;
the configuration module 1 is used for installing the identity verification equipment 2 on the terminal equipment as an edge node of a network;
the virtual training unit 31 is configured to obtain a certain number of face training samples, train the convolutional neural network of the cloud server 3 using the face training samples, and obtain the trained convolutional neural network of the cloud server 3;
the integration unit 32 is configured to integrate weights of each layer of the convolutional neural network of the cloud server 3 to obtain a weight configuration script, and send the weight configuration script to an edge node through the network;
the weight configuration unit 21 is configured to download a weight configuration script, and the weight adjustment unit 22 is configured to adjust weights of each layer of the face recognizer 23 using a convolutional neural network in the identity verification device with reference to the weight configuration script.
Specifically, the convolutional neural network comprises an input layer, a convolutional layer, a pooling layer, a full-link layer and a classification output layer.
Specifically, the specific process of training the neural network by the virtual training unit 31 is as follows: acquiring a face training sample from the input layer and transmitting the face training sample to the convolutional layer; the convolutional layer extracts the characteristics of the face training sample and transmits the characteristics to the pooling layer; the pooling layer performs pooling operation on the extracted features of the face training samples, reduces the dimensionality of feature data, and transmits the extracted features to the full-connection layer; the full-connection layer classifies the extracted features of the face training samples to obtain a classification model, transmits the classification model to the classification output layer and outputs a face recognition result; comparing the output face recognition result with the expected value to obtain the error between the output result and the expected value; and reversely transmitting the error, calculating the error of each layer, updating the weight of each layer according to the error until the output face recognition result is in an interval acceptable by an expected value, and obtaining a trained convolutional neural network capable of carrying out face recognition.
In order to further optimize the technical characteristics, the identity verification device further comprises a secondary training adjusting unit 24, wherein the secondary training adjusting unit 24 is used for performing autonomous training on the face recognizer 23 according to the face images collected on site to obtain an adjusting coefficient of the weight; on the basis of the weight value obtained by the weight value adjustment unit 22, multiplication operation is performed on the weight value and the adjustment coefficient, and secondary training adjustment of the weight value is completed.
As shown in fig. 3, an identity verification apparatus based on face edge calculation includes a weight configuration unit 21, a weight adjustment unit 22, and a face recognizer 23; wherein the content of the first and second substances,
the weight configuration unit 21 is used for downloading and storing a weight configuration script from the cloud;
the weight adjusting unit 22 is configured to adjust weights of each layer for the face recognizer 23 using a convolutional neural network in the identity verification device with reference to the weight configuration script.
In order to further optimize the technical characteristics, the system further comprises a secondary training adjusting unit 24, wherein the secondary training adjusting unit 24 is used for performing autonomous training on the face recognizer 23 according to the face image collected on site to obtain an adjusting coefficient of the weight; on the basis of the weight value obtained by the weight value adjustment unit 22, multiplication operation is performed on the weight value and the adjustment coefficient, and secondary training adjustment of the weight value is completed.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (6)
1. An identity verification method based on face edge calculation is characterized by comprising the following steps: s1, installing identity verification equipment on terminal equipment as an edge node of a network; s2, uploading a certain amount of face training samples to a cloud server, and training a convolutional neural network of the cloud server by using the face training samples to obtain the trained convolutional neural network of the cloud server; s3, integrating weights of all layers of a convolutional neural network of the cloud server to obtain a weight configuration script, and issuing the weight configuration script to an edge node through the network; s4, downloading a weight configuration script by the identity verification equipment, and adjusting weights of all layers by adopting a face recognizer of a convolutional neural network in the identity verification equipment according to the weight configuration script;
the convolutional neural network comprises an input layer, a convolutional layer, a pooling layer, a full-link layer and a classification output layer;
the specific process of training the neural network in the S2 is as follows: acquiring a face training sample from the input layer and transmitting the face training sample to the convolutional layer; the convolutional layer extracts the characteristics of the face training sample and transmits the characteristics to the pooling layer; the pooling layer performs pooling operation on the extracted features of the face training samples, reduces the dimensionality of feature data, and transmits the extracted features to the full-connection layer; the full-connection layer classifies the extracted features of the face training samples to obtain a classification model, transmits the classification model to the classification output layer and outputs a face recognition result; comparing the output face recognition result with the expected value to obtain the error between the output result and the expected value; and reversely transmitting the error, calculating the error of each layer, updating the weight of each layer according to the error until the output face recognition result is in an interval acceptable by an expected value, and obtaining a trained convolutional neural network capable of carrying out face recognition.
2. The identity verification method based on face edge calculation as claimed in claim 1, further comprising the steps of: the identity verification equipment carries out autonomous training on the face recognizer according to a face image acquired on site to obtain a weight value adjusting coefficient; and on the basis of the weight obtained by the S4 adjustment, performing multiplication operation on the weight and the adjustment coefficient to finish secondary training adjustment of the weight.
3. An identity verification system based on face edge calculation is characterized by comprising a configuration module (1), identity verification equipment (2) and a cloud server (3); the identity verification equipment (2) comprises a weight value adjusting unit (22), a weight value configuration unit (21) and a face recognizer (23); the cloud server (3) comprises a virtual training unit (31) and an integration unit (32); the configuration module (1) is used for installing the identity verification equipment (2) on terminal equipment as an edge node of a network; the virtual training unit (31) is used for acquiring a certain number of face training samples, training the convolutional neural network of the cloud server (3) by using the face training samples, and acquiring the trained convolutional neural network of the cloud server (3); the integration unit (32) is used for integrating weights of all layers of the convolutional neural network of the cloud server (3) to obtain a weight configuration script, and issuing the weight configuration script to an edge node through a network; the weight configuration unit (21) is used for downloading a weight configuration script, and the weight adjustment unit (22) is used for adjusting the weights of all layers of the face recognizer (23) which adopts a convolutional neural network in the identity verification equipment by referring to the weight configuration script;
the convolutional neural network comprises an input layer, a convolutional layer, a pooling layer, a full-link layer and a classification output layer;
the specific process of the virtual training unit (31) for training the neural network is as follows: acquiring a face training sample from the input layer and transmitting the face training sample to the convolutional layer; the convolutional layer extracts the characteristics of the face training sample and transmits the characteristics to the pooling layer; the pooling layer performs pooling operation on the extracted features of the face training samples, reduces the dimensionality of feature data, and transmits the extracted features to the full-connection layer; the full-connection layer classifies the extracted features of the face training samples to obtain a classification model, transmits the classification model to the classification output layer and outputs a face recognition result; comparing the output face recognition result with the expected value to obtain the error between the output result and the expected value; and reversely transmitting the error, calculating the error of each layer, updating the weight of each layer according to the error until the output face recognition result is in an interval acceptable by an expected value, and obtaining a trained convolutional neural network capable of carrying out face recognition.
4. The identity verification system based on face edge calculation as claimed in claim 3, wherein the identity verification device further comprises a secondary training adjustment unit (24), the secondary training adjustment unit (24) is used for performing autonomous training on the face recognizer (23) according to the face image collected on site to obtain the adjustment coefficient of the weight; on the basis of the weight value obtained by the weight value adjusting unit (22), multiplication operation is carried out on the weight value and the adjusting coefficient, and secondary training adjustment of the weight value is completed.
5. An identity verification device based on face edge calculation is characterized by comprising a weight value configuration unit (21), a weight value adjusting unit (22) and a face recognizer (23); the weight value configuration unit (21) is used for downloading and storing a weight value configuration script from a cloud end; the weight value adjusting unit (22) is used for adjusting the weight values of all layers of the face recognizer (23) which adopts a convolutional neural network in the identity verification equipment by referring to the weight value configuration script;
the convolutional neural network comprises an input layer, a convolutional layer, a pooling layer, a full-link layer and a classification output layer;
the specific training process of the convolutional neural network comprises the following steps: acquiring a face training sample from the input layer and transmitting the face training sample to the convolutional layer; the convolutional layer extracts the characteristics of the face training sample and transmits the characteristics to the pooling layer; the pooling layer performs pooling operation on the extracted features of the face training samples, reduces the dimensionality of feature data, and transmits the extracted features to the full-connection layer; the full-connection layer classifies the extracted features of the face training samples to obtain a classification model, transmits the classification model to the classification output layer and outputs a face recognition result; comparing the output face recognition result with the expected value to obtain the error between the output result and the expected value; and reversely transmitting the error, calculating the error of each layer, updating the weight of each layer according to the error until the output face recognition result is in an interval acceptable by an expected value, and obtaining a trained convolutional neural network capable of carrying out face recognition.
6. The identity verification device based on face edge calculation as claimed in claim 5, further comprising a secondary training adjustment unit (24), wherein the secondary training adjustment unit (24) is configured to perform autonomous training on the face recognizer (23) according to a face image acquired in the field, and obtain an adjustment coefficient of a weight; on the basis of the weight value obtained by the weight value adjusting unit (22), multiplication operation is carried out on the weight value and the adjusting coefficient, and secondary training adjustment of the weight value is completed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010078951.6A CN111339833B (en) | 2020-02-03 | 2020-02-03 | Identity verification method, system and equipment based on face edge calculation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010078951.6A CN111339833B (en) | 2020-02-03 | 2020-02-03 | Identity verification method, system and equipment based on face edge calculation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111339833A CN111339833A (en) | 2020-06-26 |
| CN111339833B true CN111339833B (en) | 2022-10-28 |
Family
ID=71185161
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010078951.6A Active CN111339833B (en) | 2020-02-03 | 2020-02-03 | Identity verification method, system and equipment based on face edge calculation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111339833B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106650699A (en) * | 2016-12-30 | 2017-05-10 | 中国科学院深圳先进技术研究院 | CNN-based face detection method and device |
| CN108062533A (en) * | 2017-12-28 | 2018-05-22 | 北京达佳互联信息技术有限公司 | Analytic method, system and the mobile terminal of user's limb action |
| CN108073917A (en) * | 2018-01-24 | 2018-05-25 | 燕山大学 | A kind of face identification method based on convolutional neural networks |
| CN109522872A (en) * | 2018-12-04 | 2019-03-26 | 西安电子科技大学 | A kind of face identification method, device, computer equipment and storage medium |
| CN109523461A (en) * | 2018-11-09 | 2019-03-26 | 北京达佳互联信息技术有限公司 | Method, apparatus, terminal and the storage medium of displaying target image |
| CN110163169A (en) * | 2019-05-27 | 2019-08-23 | 北京达佳互联信息技术有限公司 | Face identification method, device, electronic equipment and storage medium |
| CN110516649A (en) * | 2019-09-02 | 2019-11-29 | 南京微小宝信息技术有限公司 | Alumnus's authentication method and system based on recognition of face |
| CN110633648A (en) * | 2019-08-21 | 2019-12-31 | 重庆特斯联智慧科技股份有限公司 | Face recognition method and system in natural walking state |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8418206B2 (en) * | 2007-03-22 | 2013-04-09 | United Video Properties, Inc. | User defined rules for assigning destinations of content |
-
2020
- 2020-02-03 CN CN202010078951.6A patent/CN111339833B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106650699A (en) * | 2016-12-30 | 2017-05-10 | 中国科学院深圳先进技术研究院 | CNN-based face detection method and device |
| CN108062533A (en) * | 2017-12-28 | 2018-05-22 | 北京达佳互联信息技术有限公司 | Analytic method, system and the mobile terminal of user's limb action |
| CN108073917A (en) * | 2018-01-24 | 2018-05-25 | 燕山大学 | A kind of face identification method based on convolutional neural networks |
| CN109523461A (en) * | 2018-11-09 | 2019-03-26 | 北京达佳互联信息技术有限公司 | Method, apparatus, terminal and the storage medium of displaying target image |
| CN109522872A (en) * | 2018-12-04 | 2019-03-26 | 西安电子科技大学 | A kind of face identification method, device, computer equipment and storage medium |
| CN110163169A (en) * | 2019-05-27 | 2019-08-23 | 北京达佳互联信息技术有限公司 | Face identification method, device, electronic equipment and storage medium |
| CN110633648A (en) * | 2019-08-21 | 2019-12-31 | 重庆特斯联智慧科技股份有限公司 | Face recognition method and system in natural walking state |
| CN110516649A (en) * | 2019-09-02 | 2019-11-29 | 南京微小宝信息技术有限公司 | Alumnus's authentication method and system based on recognition of face |
Non-Patent Citations (3)
| Title |
|---|
| Face recognition in video streams for mobile assistive devices dedicated to visually impaired;Ruxandra Tapu等;《网页在线公开：https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8706187》;20190506;第1-6页 * |
| 人脸边缘表示与估计的识别方法研究;侯景;《科技信息》;20140527(第15期);第2912-2929页 * |
| 基于差值局部方向模式的人脸特征表示;李照奎等;《软件学报》;20160819;第26卷(第11期);第64-65页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111339833A (en) | 2020-06-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109359636B (en) | Video classification method, device and server | |
| CN107977390B (en) | Method, device and system for identifying picture book and electronic equipment | |
| CN110147710B (en) | Method and device for processing human face features and storage medium | |
| US20210160556A1 (en) | Method for enhancing resolution of streaming file | |
| WO2019015682A1 (en) | Dynamic facial image warehousing method and apparatus, electronic device, medium, and program | |
| CN109299709A (en) | Data recommendation method, device, server end and client based on recognition of face | |
| Zhang et al. | Perceptual generative autoencoders | |
| US20140095536A1 (en) | System and method for analysis of muitimedia content items | |
| CN111049659B (en) | Service verification method, device and system based on handwriting signature recognition | |
| US20230326468A1 (en) | Audio processing of missing audio information | |
| CN112257612B (en) | Unmanned aerial vehicle video frame filtering method and device based on edge intelligence | |
| CN110796000B (en) | Bidirectional LSTM-based lip sample generation method, device and storage medium | |
| CN111541632A (en) | Physical layer authentication method based on principal component analysis and residual error network | |
| CN113936298A (en) | Feature recognition method and device and computer readable storage medium | |
| CN112491442A (en) | Self-interference elimination method and device | |
| CN113191461A (en) | Picture identification method, device and equipment and readable storage medium | |
| CN111339833B (en) | Identity verification method, system and equipment based on face edge calculation | |
| WO2022246986A1 (en) | Data processing method, apparatus and device, and computer-readable storage medium | |
| CN108683901B (en) | Data processing method, MEC server and computer readable storage medium | |
| KR102130076B1 (en) | Method for improving the resolution of streaming files based on the learning importance of feature areas | |
| CN112926436A (en) | Behavior recognition method and apparatus, electronic device, and storage medium | |
| KR20220058915A (en) | Image detection and related model training methods, apparatus, apparatus, media and programs | |
| CN113313086B (en) | Feature vector conversion model processing method, device, server and storage medium | |
| CN113065579B (en) | Method and device for classifying target object | |
| CN115359390A (en) | Image processing method and device |
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 |