WO2019090769A1

WO2019090769A1 - Human face shape recognition method and apparatus, and intelligent terminal

Info

Publication number: WO2019090769A1
Application number: PCT/CN2017/110711
Authority: WO
Inventors: 林丽梅
Original assignee: 深圳和而泰智能控制股份有限公司
Priority date: 2017-11-13
Filing date: 2017-11-13
Publication date: 2019-05-16
Also published as: CN108701216B; CN108701216A

Abstract

Provided are a human face shape recognition method, an intelligent terminal and a storage medium. The method comprises: acquiring a human face image; extracting a human face key point from the human face image; based on the human face key point, determining a reference value of the human face image; building a feature vector of the human face image by combining the human face key point with the reference value; inputting the feature vector into a face shape feature model; and acquiring a face shape recognition result output by the face shape feature model. By means of the technical solution, the embodiments of the present application can recognize a human face shape by combining human face key points other than a human face outline and by combining multiple human face features, thereby improving the reliability of a face shape recognition result.

Description

Face face recognition method, device and intelligent terminal

Technical field

The present application relates to the field of face recognition technologies, and in particular, to a face face recognition method, apparatus, and smart terminal.

Background technique

Face recognition technology is a technology for identifying and comparing facial visual feature information. Its research fields include: identification, expression recognition, gender recognition and beauty skin care.

In recent years, with the increasing material living standards, people's demand for personal image design has grown rapidly. To provide users with a personal image design usually needs to first determine the user's face, and then select the appropriate hairstyle, makeup, glasses, accessories, etc. according to the user's face. Based on this demand, the field of face recognition technology has also proposed some methods for detecting a person's face by recognizing a face in an image. For example, in the prior art, the face shape of the face to be tested can be identified by first positioning the contour of the face and then using the curvature feature of the contour of the face.

However, in the process of implementing the present application, the inventors have found that the prior art has at least the following problems: the recognition of the face face based on the curvature feature of the face contour requires a higher recognition accuracy of the face contour curve, but in practice In the application, the extraction of the mandible contour curve is difficult, and the accuracy of the extracted contour curve is not high. If the face feature based on the curvature feature of the face contour is used to recognize the face shape, the face face recognition result is low in reliability.

Summary of the invention

The embodiment of the present invention provides a face face recognition method, device, and intelligent terminal, which can solve the problem of low reliability of the recognition result of the face recognition based on the curvature of the face contour.

In a first aspect, the embodiment of the present application provides a method for recognizing a face face, including:

Obtaining a face image;

Extracting a face key point in the face image, wherein the face key point includes: a nasal bone off Key points, key points of the mandible, and key points of the chin;

Determining a reference value of the face image based on the face key point;

Constructing a feature vector of the face image in combination with the face key point and the reference value, wherein the feature vector includes: a face length feature value, a mandible width feature value, and a chin angle feature value, the person A face length feature value is constructed based on the nasal bone key point, the chin key point, and the reference value, the mandible width feature value being constructed based on the mandible key point and the reference value, the chin angle feature value Constructing based on the key points of the mandible and the key points of the chin;

Inputting the feature vector into a face feature model;

Obtaining a face recognition result output by the face feature model.

Optionally, the face key point further includes two axillary key points;

Then, determining, according to the face key point, the reference value of the face image, including:

The distance between the two axillary key points is used as a reference value of the face image.

Optionally, before the step of inputting the feature vector into the face feature model, the face face recognition method further includes:

Obtaining the face feature model.

Optionally, the acquiring the face feature model includes:

Collecting a preset number of face image samples, each of the face image samples being marked with a face mark;

Extracting face key points in the face image sample respectively, wherein the face key points include: a key point of the nose bone, a key point of the mandible, and a key point of the chin;

Determining a reference value of each of the face image samples based on the face key points;

Constructing a feature vector of each of the face image samples in combination with the face key point and the reference value, wherein the feature vector includes: a face length feature value, a mandible width feature value, and a chin angle feature value. The face length feature value is constructed based on the nasal bone key point, the chin key point, and the reference value, the mandible width feature value is based on the mandible key point and the reference value Constructing, the chin angle feature value is constructed based on the key point of the mandible and the key point of the chin;

The feature vector and the face mark of each of the face image samples are input into a support vector machine model, and the face feature model is trained.

Optionally, the face key point further includes: a tibia key point, the feature vector further includes: a cheek width feature value, the cheek width feature value is based on the nasal bone key point, the tibia key point, and The reference value is constructed.

Optionally, the feature vector further includes: a side face length feature value, and the side face length feature value is constructed based on the mandible key point, the tibial key point, and the reference value.

Optionally, the feature vector further includes: a ratio of the side face length feature value to the face length feature value.

In a second aspect, the embodiment of the present application provides a face recognition device, including:

An image acquisition unit, configured to acquire a face image;

a face key point extracting unit, configured to extract a face key point in the face image, wherein the face key point comprises: a key point of a nose bone, a key point of a mandible, and a key point of a chin;

a reference value determining unit, configured to determine a reference value of the face image based on the face key point;

a feature vector construction unit, configured to construct a feature vector of the face image in combination with the face key point and the reference value, wherein the feature vector includes: a face length feature value, a mandible width feature value, and a chin An angle feature value, the face length feature value being constructed based on the nasal bone key point, the chin key point, and the reference value, the mandible width feature value being constructed based on the mandible key point and the reference value The chin angle feature value is constructed based on the key point of the mandible and the key point of the chin;

a matching unit, configured to input the feature vector into a face feature model;

And an output unit, configured to acquire a face recognition result output by the face feature model.

Optionally, the facial face recognition device further includes:

a model obtaining unit, configured to acquire the face feature model.

Optionally, the model obtaining unit includes:

a face image sample collection module, configured to collect a preset number of face image samples, each of the face image samples being marked with a face mark;

a face key point extraction module, configured to separately extract face key points in the face image sample, wherein the face key points include: a key point of a nose bone, a key point of a mandible, and a key point of a chin;

a reference value determining module, configured to determine a reference value of each of the face image samples based on the face key points respectively;

a feature vector construction module, configured to construct a feature vector of each of the face image samples in combination with the face key point and the reference value, wherein the feature vector comprises: a face length feature value, a mandible width An eigenvalue and a chin angle feature value, the face length feature value being constructed based on the nasal bone key point, the chin key point, and the reference value, the mandible width feature value being based on the mandible key point and Constructing a reference value, the chin angle feature value being constructed based on the mandible key point and the chin key point;

And a training module, configured to input a feature vector and a face mark of each of the face image samples into a support vector machine model, and train the face feature model.

In a third aspect, an embodiment of the present application provides an intelligent terminal, including:

At least one processor; and,

a memory communicatively coupled to the at least one processor; wherein

The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the face face recognition method as described above.

In a fourth aspect, the embodiment of the present application provides a storage medium, where the storage medium stores executable instructions, and when the executable instructions are executed by the smart terminal, the smart terminal is configured to perform the face face recognition method as described above. .

In a fifth aspect, an embodiment of the present application further provides a program product, where the program product includes a program stored on a storage medium, where the program includes program instructions, when the program instructions are executed by the smart terminal, The smart terminal performs the face face recognition method as described above.

The beneficial effects of the embodiment of the present application are: a face face recognition method provided by the embodiment of the present application, The device and the smart terminal extract a face key point in the face image when acquiring a face image, determine a reference value of the face image based on the face key point, and combine the face key point Constructing a feature vector of the face image with the reference value, wherein the face key point comprises: a nose bone key point, a mandible key point, and a chin key point, and the feature vector includes: a face length feature value, a mandible width feature value and a chin angle feature value, the mandible width feature value being constructed based on the mandible key point and the reference value, the chin angle feature value being based on the mandible key point and the chin key Point construction; then input the feature vector into the face feature model; finally obtain the face recognition result output by the face feature model, and can combine the face key points outside the face contour, for example, the key points of the nose bone, combined with various people The face feature recognizes the face face type, thereby improving the reliability of the face recognition result.

DRAWINGS

The one or more embodiments are exemplified by the accompanying drawings in the accompanying drawings, and FIG. The figures in the drawings do not constitute a scale limitation unless otherwise stated.

1 is a schematic flowchart of a face face recognition method according to an embodiment of the present application;

2 is a schematic diagram of an example of a face key point location provided by an embodiment of the present application;

3 is a schematic flowchart of a method for acquiring a face feature model according to an embodiment of the present application;

4 is a schematic diagram of an example of a face face classification provided by an embodiment of the present application;

FIG. 5 is a schematic structural diagram of a face face recognition device according to an embodiment of the present application; FIG.

FIG. 6 is a schematic structural diagram of hardware of an intelligent terminal according to an embodiment of the present application.

Detailed ways

In order to make the objects, technical solutions, and advantages of the present application more comprehensible, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are merely illustrative of the application and are not intended to be limiting.

It should be noted that, if there is no conflict, the various features in the embodiments of the present application may be combined with each other, and are all within the protection scope of the present application. In addition, although the functional module division is performed in the device schematic, The logical order is shown in the flowcharts, but in some cases the steps shown or described may be performed in a different order than the modules in the device, or in the sequence of the flowchart.

The embodiment of the present application provides a face face recognition method, device, intelligent terminal, and storage medium. The face face recognition method is a face recognition method based on a feature vector constructed by a face key point and a face feature model trained by a machine learning algorithm, and the face is extracted when the face image is acquired. a face key point in the image, determining a reference value of the face image based on the face key point, and constructing a feature vector of the face image according to the face key point and the reference value, where The key points of the face include: a key point of the nose bone, a key point of the mandible, and a key point of the chin, and the feature vector includes: a face length feature value, a mandible width feature value, and a chin angle feature value, and the mandible width feature value Constructed based on the mandible key point and the reference value, the chin angle feature value is constructed based on the mandible key point and the chin key point; then the feature vector is input into the face feature model; and finally the The face recognition result output by the face feature model can be combined with face key points outside the face contour, for example, key points of the nose bone, combined with various people Facial features to recognize faces, facial recognition so as to enhance reliability of the results.

The face face recognition method, the smart terminal and the storage medium provided by the embodiments of the present application can be applied to any technical field related to face recognition, such as portrait nationality recognition, etc., and are particularly suitable for the fields of beauty application, personal image design and the like. For example, a beauty-like application can be developed based on the inventive concept of the face face recognition method provided by the embodiment of the present application, and the application can automatically recognize the face shape of the face image when the user inputs the face image, and further Faces are designed with appropriate hair styles, makeup, eyeglass frames, jewelry, and more.

The face face recognition method provided by the embodiment of the present application may be performed by any type of smart terminal having an image processing function, and the smart terminal may include any suitable type of storage medium for storing data, such as a magnetic disk or a compact disk (CD). -ROM), read-only memory or random storage memory. The smart terminal may also include one or more logical computing modules that perform any suitable type of function or operation in parallel, such as viewing a database, image processing, etc., in a single thread or multiple threads. The logic operation module may be any suitable type of electronic circuit or chip-type electronic device capable of performing logical operation operations, such as a single core processor, a multi-core processor, a graphics processing unit (GPU), or the like. Lift For example, the smart terminal may include, but is not limited to, a beauty authentication device, a personal computer, a tablet computer, a smart phone, a server, and the like.

Specifically, the embodiments of the present application are further described below in conjunction with the accompanying drawings.

1 is a schematic flowchart of a face face recognition method according to an embodiment of the present application. Referring to FIG. 1, the method includes but is not limited to the following steps:

Step 110: Acquire a face image.

In the present embodiment, the "face image" refers to an image including the face of the detected person, by which all face features of the detected person can be acquired.

In this embodiment, the specific implementation manner of acquiring the face image may be: collecting the positive face image of the detected person in real time; or, directly: acquiring the existing detected person including the detected person directly in the smart terminal or in the cloud. The image of the face. For different application scenarios or the selection of the detected person, different ways of acquiring the face image may be selected. For example, it is assumed that a smart terminal for recommending a suitable frame for the user is provided in the eyeglass store, and in order to be able to recommend a suitable frame based on the face of the user in time, the manner in which the smart terminal acquires the face image may be by camera. The device collects the positive face image of the detected person in real time. For example, the user wants to design a suitable makeup for himself through his own smart terminal, for example, a smart phone. Since the smart terminal generally stores a personal face image, the smart terminal acquires the face in the application scenario. The image may also be obtained by directly capturing the existing image including the positive face of the detected person directly in the smart terminal or in the cloud. Of course, in the actual application, the manner of obtaining the face image is not limited to the above description, and the comparison of the embodiments of the present application is not specifically limited.

Step 120: Extract a face key point in the face image.

In this embodiment, the face key point refers to a feature point distributed on an area having a characteristic feature in a face (for example, a face contour, an eye, an eyebrow, a nose, a mouth, etc.). Different faces have different face key points. Wherein, each of the regions having the characteristic features may be distributed with a plurality of face key points. For example, in the nose region, a key point of the nose for identifying the position of the bridge of the nose, a key point of the nose for identifying the position of the nose, and the like may be included; for example, on the contour area of the face, may include a marker for identifying the mandible. The contoured mandible key points, the humeral key points used to identify the contour of the face near the tibia, the chin key points used to identify the position of the chin, and the like. In addition, it should be noted that In this embodiment, the number of key points of the nose, the key points of the nose, the key points of the mandible, the key points of the tibia, and the key points of the chin are not limited to one.

In this embodiment, the acquired face image may be positioned by using a suitable algorithm or manner, and then the required face key points are extracted. In particular, in the present embodiment, in order to accurately describe the face feature of the face image, the key points that need to be extracted include, but are not limited to, a key point of the nose, a key point of the mandible, and a key point of the chin.

For example, the face key point of the obtained face image can be located through the third-party toolkit dlib, and the obtained key point distribution map is shown in FIG. 2 (where the face key points 0-16 are used to identify the person) The feature points of the face contour area of the face image, the face key points 27-30 are the feature points of the nasal bone area identifying the face image); then the preset face key points are selected and the coordinate parameters are obtained, for example, Select the key points 27 to 30 of the nasal bone, which are located at the key points 4-7 and 9-12 of the mandible on both sides of the face, and the key point 8 of the chin. It can be understood that, in this embodiment, for the exemplary description, only the face contour and the face key point of the nasal bone region are marked, and in actual application, the face obtained by the third party tool package dilb is positioned. The key points can include 56.

In addition, it can be understood that, in practical applications, in order to unify the standard, the same face key point positioning method is adopted for different face images, and the same face key points are extracted for subsequent analysis.

Step 130: Determine a reference value of the face image based on the face key point.

In practical applications, when the person takes a picture, the distance from the camera device is different. Therefore, in the collected face image, there may be some face occupying a large screen area, and some faces occupy a small screen area. . Therefore, in order to make the face images having different face occupying areas have comparability, in the embodiment, a reference standard is set for the facial feature values of each face image, and the reference standard is the reference. value".

Specifically, in this embodiment, the distance between any two extracted face key points may be used as a reference value of the face image, for example, the distance between the nose bone key point 27 and the nose bone key point 30 is taken as The reference value of the face image; or, the distance between the following key point 4 of the jaw and the key point 12 of the mandible is used as a reference value of the face image; or, the following key points of the jaw bone 5 and the key point of the chin 8 The distance between them is used as a reference value for the face image. Among them, it should be noted that, in order to unify the standard, when a method of constructing a reference value is selected, each face image constructs its reference value according to the construction method. For example, suppose the distance between the key point 4 of the jaw and the key point 12 of the mandible is selected to construct a reference value. If the face image A is acquired at this time, the key point 4 of the mandible of the face image A is extracted. The key point 12 of the mandible is determined, and the reference value of the face image A is determined as the distance between the key point 4 of the mandible image A and the key point 12 of the mandible; if the face image B is acquired at this time, the person is extracted The mandibular key point 4 of the face image B and the mandible key point 12, and the reference value of the face image B is determined as the distance between the mandible key point 4 of the face image B and the mandible key point 12.

Wherein, in some embodiments, the extracted face key points further include two armpit key points located at the temple position, that is, face

key points

1 and 15 as shown in FIG. 2, at this time, based on the person A specific embodiment of determining a reference value of the face image by the face key may be: using a distance between the two axillary key points as a reference value of the face image. Wherein, since the distance between the key points of the two armpits is a major feature of the face type, the distance is used as the reference value of the face image to obtain the distance between the key points of the specific face (for example, located on both sides of the face) The ratio of the distance between the two key points of the mandible and the distance can better reflect the facial features of the face image.

Step 140: Construct a feature vector of the face image in combination with the face key point and the reference value.

In the present embodiment, the "feature vector" is a parameter for characterizing a face face of the acquired face image, which may be composed of facial feature values of a plurality of face images. The “facial feature value” is a parameter that reflects a facial feature of the face image, and may include, but is not limited to, a face length feature value, a mandible width feature value, and a chin angle feature value.

Specifically, in this embodiment, a plurality of preset facial feature values may be first constructed based on the extracted face key points and the determined reference values, such as: based on the nasal bone key points, the chin key points, and The reference value constructs a face length feature value of the face image, and constructs a mandible width feature value of the face image based on the mandible key point and the reference value, and based on the key point of the mandible The chin key points construct a chin angle feature value of the face image; and then the constructed face feature values are combined into a feature vector of the face image.

In practical applications, in order to improve the accuracy of facial face recognition, more types of facial feature values related to face types can be constructed based on a larger number of face key points, and the number of facial feature values of each type can also be Not limited to one. For example, in some embodiments, the extracted face key points may further include a key point of the humerus on the contour of the face near the tibia, which may be based on the extracted key points of the nasal bone and the key points of the tibia ( And/or, the critical point of the mandible and the determined reference value to construct a cheek width feature value of the face image, and/or based on the extracted key points of the mandible, key points of the tibia and the determined reference values The side face length feature value of the face image. Further, in still other embodiments, the constructed feature vector may further include a ratio of any two facial feature values having length/width features, such as the side face length feature value and the face length feature value. The ratio, the ratio of the side face length feature value to the mandible width feature value, and the like.

For example: suppose the obtained key point distribution is as shown in Figure 2, and the distance between two armpit key points (face key points 1 and 15) is used as the reference value of the face image. The manner of constructing the feature vector of the face image in combination with the face key point and the reference value may be:

First, the following 18 facial feature values are constructed in conjunction with the face key point in FIG. 2 and the reference value D(1, 15):

Feature1: D(8,27)/D(1,15);

Feature2: D(4,12)/D(1,15);

Feature3: D(5,11)/D(1,15);

Feature4: D(6,10)/D(1,15);

Feature5: D(9,10)/D(1,15);

Feature6: D(8,12)/D(1,15);

Feature7: D(12,14)/D(1,15);

Feature8: Feature3/Feature1;

Feature9: Feature7/Feature1;

Feature10: Feature7/Feature3;

Feature11: (D(27,3)+D(27,13))/2/D(1,15);

Feature12: (D(27,4)+D(27,12))/2/D(1,15);

Feature13: (D(27,5)+D(27,11))/2/D(1,15);

Feature14: (D(27,6)+D(27,10))/2/D(1,15);

Feature15: the cosine of the angle formed by L(8,5) and L(8,11);

Feature16: the cosine of the angle formed by L(8,6) and L(8,10);

Feature17: the cosine of the angle formed by L(8,7) and L(8,9);

Feature18: the cosine of the angle formed by L(11,10) and L(11,12);

Where "/" indicates the division number, D(x, y) represents the distance between the face key point x and the face key point y, and L(x, y) represents the face key point x and the face key point y The line segment between.

Among them, Feature1 is the face length feature value; Feature2~Feature4 is the mandible width feature value; Feature5 and Feature6 are the mandible length feature values; Feature7 is the lateral face length feature value; Feature8 is the mandible width feature value and face length feature value. The ratio of Feature9 is the ratio of the length value of the face length to the feature value of the face length; Feature10 is the ratio of the feature value of the face length to the feature value of the mandible width; Feature11~Feature14 are the feature values of the cheek width; Feature15~Feature18 are the chin angle. Eigenvalues.

Then, the feature vectors of the face image are composed of these facial feature values, namely: Face=[Feature1, Feature2, Feature3Feature4, Feature5, Feature6, Feature7, Feature8, Feature9, Feature10, Feature11, Feature12, Feature13, Feature14, Feature15, Feature16, Feature 17, Feature 18], this feature is a numerical indication of the facial features of the face image.

Step 150: Input the feature vector into the face feature model.

In this embodiment, the "face feature model" is a face type classifier trained by a machine learning algorithm, and after inputting the feature vector of the face image into the face feature model, the face feature model can input the feature vector An operation is performed to output a face recognition result corresponding to the feature vector. The face feature model can be pre-trained and stored locally in the smart terminal. When the face face recognition is performed, the face feature model can be directly called from the smart terminal.

In some embodiments, the smart terminal does not pre-store the face feature model. In this embodiment, before performing step 150, the face feature model needs to be acquired first. The specific embodiment manner of obtaining the face feature model may be: connecting from other devices or the cloud through a network connection. The face feature model is carried; or the face feature model can be trained based on the face image sample and the machine learning algorithm locally on the smart terminal.

The specific implementation manner of training the face feature model based on the face image sample and the machine learning algorithm may be: first collecting a predetermined number of face image samples; and then extracting a feature vector of each face image sample as a training sample; Finally, any suitable machine learning algorithm, such as neural network, decision tree, support vector machine, etc., is used to train the training samples, so that the trained model has the function of face face classification.

Specifically, as shown in FIG. 3, a schematic flowchart of a method for acquiring the face feature model provided by the embodiment of the present application may include, but is not limited to, the following steps:

Step 151: Collect a preset number of face image samples, and each of the face image samples is marked with a face mark.

In an embodiment, the face face is first classified according to needs, for example, the face face is divided into five categories: a heart-shaped face, a goose face, a long face, a round face, and a square face as shown in FIG. 4; The face shape collects a predetermined number of face image samples, and a face type mark is marked in each face image sample for indicating the face type to which the face image sample belongs. For example, collecting 100 face image samples with heart-shaped faces and labeling these face image samples with a face mark "1" (used to represent a heart-shaped face); collecting 100 face image samples with a goose face and These face image samples are marked with a face mark "2" (used to indicate that it is a goose face); 100 face image samples with long faces are collected and face type marks "3" are marked for these face image samples (used to indicate that they are long faces) Collecting 100 face image samples with round faces and labeling these face image samples with face mark "4" (used to represent a round face); 100 face image samples with square faces for these people The face image sample is marked with a face mark "5" (used to indicate that it is a square face). Thereby, 500 face image samples can be obtained, and each face image sample is marked with a face mark for indicating the face type to which it belongs.

Step 152: Extract the face key points in the face image sample separately.

In this embodiment, the same face key point positioning method is used, for example, the third party toolkit dilb is used to perform face key point positioning on all collected face image samples and extract the predetermined position. Face key points, including but not limited to: key points of the nasal bone, key points of the mandible, and key points of the chin. For the specific implementation manner of extracting the key points of the face image in each face image sample, refer to step 120 above, and details are not described herein again.

Among them, it can be understood that, in order to unify the standard, in the face face recognition, the same face key point positioning method is also used to perform face key point positioning on the acquired face image and extract the same face key point. .

Step 153: Determine a reference value of each of the face image samples based on the face key points, respectively.

In this embodiment, the step 153 has the same technical features as the above-mentioned step 130, and the specific implementation manner can also refer to the above step 130, and therefore, details are not described herein again.

Among them, it can be understood that, in order to unify the standard, when training the face feature model and performing face face recognition, the same method is needed to determine the reference value of each face image sample or face image, for example, both The distance between the key points of the armpit is used as the reference value.

Step 154: Construct a feature vector of each of the face image samples in combination with the face key point and the reference value, respectively.

In the embodiment, the step 154 has the same technical features as the above-mentioned step 140, and the specific embodiment can also refer to the above step 140, and therefore, details are not described herein again.

Among them, it can be understood that, in order to unify the standard, when training the face feature model and performing face face recognition, it is also necessary to construct the feature vector of each face image sample or face image in the same manner.

Step 155: Input feature vectors and face markers of each of the face image samples into a training support vector machine model, and train the face feature model.

In this embodiment, based on the face markers and feature vectors of all face image samples, a face vector feature model with face face classification function is trained by using a Support Vector Machine (SVM) model. Among them, the "Support Vector Machine Model" is a supervised machine learning model, which is usually used for pattern recognition, classification and regression analysis.

Specifically, in the embodiment, the feature vector and the face mark of each of the face image samples are The support vector machine model is input, and the face feature model is trained. Specifically, the feature vector of each face image sample is input as a variable, and the face mark of the face image sample is input as a result; through a large number of training samples, a function describing the face shape can be obtained, which is equivalent to training generation. Face feature model. Thus, when a variable input (feature vector of the face image to be tested) is received, a result (ie, a face recognition result of the face image to be tested) can be obtained by the function (face type feature model).

In practical applications, the machine learning toolkit sklearn can be called and the svm class can be introduced to generate an empty object model; then the training data (the feature vector of each face image sample and its face type mark) is fed to the model.fit; The model.fit can be trained to obtain a face feature model.

Further, the parameters of the trained face feature model may also be saved locally in the smart terminal, so as to facilitate the face face recognition to be directly called when the facial face recognition is performed later.

Step 160: Acquire a face recognition result output by the face feature model.

In this embodiment, by calculating the input variable, that is, the feature vector of the face image, the face type recognition result is obtained, and the face recognition result is the face in the face image. Face shape.

According to the foregoing technical solution, the human face recognition method provided by the embodiment of the present application extracts a face key point in the face image by acquiring a face image. The face key points determine a reference value of the face image and construct a feature vector of the face image in combination with the face key point and the reference value, wherein the face key point comprises: a nose bone key point a key point of the mandible and a key point of the chin, the feature vector comprising: a face length feature value, a mandible width feature value, and a chin angle feature value, the mandible width feature value being based on the mandible key point and the a reference value construction, the chin angle feature value is constructed based on the mandible key point and the chin key point; then the feature vector is input into the face feature model; and finally the face recognition result output by the face feature model is obtained, Combine face key points outside the contour of the face, such as key points of the nose, combined with a variety of facial features to recognize the face face, thereby enhancing the face The reliability of the recognition result.

FIG. 5 is a schematic structural diagram of a face face recognition device according to an embodiment of the present disclosure. Referring to FIG. 5, the face recognition device 5 includes but is not limited to:

An image obtaining unit 51, configured to acquire a face image;

a face key extraction unit 52, configured to extract a face key point in the face image, wherein the face key point includes: a key point of the nose bone, a key point of the mandible, and a key point of the chin;

a reference value determining unit 53 configured to determine a reference value of the face image based on the face key point;

The feature vector construction unit 54 is configured to construct a feature vector of the face image in combination with the face key point and the reference value, wherein the feature vector includes: a face length feature value, a mandible width feature value, and a chin angle feature value, the face length feature value being constructed based on the nasal bone key point, the chin key point, and the reference value, the mandible width feature value being based on the mandible key point and the reference value Constructing, the chin angle feature value is constructed based on the key point of the mandible and the key point of the chin;

a matching unit 55, configured to input the feature vector into a face feature model; and,

The output unit 56 is configured to acquire a face recognition result output by the face feature model.

In the embodiment of the present application, when the image acquisition unit 51 acquires the face image, the face key point extraction unit 52 determines the reference value of the face image based on the face key point; Constructing a feature vector of the face image in combination with the face key point and the reference value; then inputting the feature vector into the face feature model by the matching unit 55; finally, acquiring the face feature by the output unit 56 The face recognition result of the model output.

In some embodiments, the extracted face key points further include: two armpit key points; at this time, the reference value determining unit 53 is specifically configured to use the distance between the two axillary key points as a The reference value of the face image.

In some embodiments, the face face recognition device 5 further includes: a model acquisition unit 57, configured to acquire the face feature model.

Specifically, in some of the embodiments, the model obtaining unit 57 includes: a face image sample collecting module 571, a face key point extracting module 572, a reference value determining module 573, and a feature vector building model. Block 574 and training module 575. The face image sample collection module 571 is configured to collect a preset number of face image samples, each of the face image samples is labeled with a face mark; and the face key point extraction module 572 is configured to separately extract the face image. a face key point in the sample, wherein the face key point includes: a nasal bone key point, a mandibial key point, and a chin key point; the reference value determining module 573 is configured to determine each of the face key points respectively a reference value of the face image sample; the feature vector construction module 574 is configured to construct a feature vector of each of the face image samples in combination with the face key point and the reference value, wherein the feature vector comprises: a face length feature value, a mandible width feature value, and a chin angle feature value, the face length feature value being constructed based on the nasal bone key point, the chin key point, and the reference value, the mandible width feature value Constructed based on the mandible key point and the reference value, the chin angle feature value is constructed based on the mandible key point and the chin key point; training module 5 75 is configured to input feature vectors and face markers of each of the face image samples into a support vector machine model, and train the face feature model.

In addition, in some embodiments, the face key point may further include: a tibia key point, and the feature vector may further include: a cheek width feature value, and/or a side face length feature value. The cheek width feature value is constructed based on the nasal bone key point, the tibia key point, and the reference value; the side face length feature value is based on the mandible key point, the tibia key point, and the Baseline value build. Further, in still other embodiments, the feature vector may further include a ratio of the side face length feature value to the face length feature value. It can be understood that, in practical applications, a larger number of face-related feature values can be constructed by constructing a larger number of face-related feature values to construct feature vectors of face images, which are not enumerated here.

It should be noted that, since the face face recognition device and the face face recognition method in the above method embodiment are based on the same inventive concept, the corresponding content of the above method embodiment is also applicable to the device embodiment. It will not be detailed.

The above-mentioned technical solutions can be seen that the facial face recognition device provided by the embodiment of the present application is based on the face key point extraction unit 52 by the face key point extraction unit 52 when the image acquisition unit 51 acquires the face image. The face key points determine a reference value of the face image; and further, the feature vector constructing unit 54 combines the face key point and the reference value to construct the feature of the face image a vector, wherein the face key points include: a nasal bone key point, a mandibial key point, and a chin key point, the feature vector including: a face length feature value, a mandible width feature value, and a chin angle feature value, A mandible width feature value is constructed based on the mandible key point and the reference value, the chin angle feature value being constructed based on the mandible key point and the chin key point; the feature vector is then passed by the matching unit 55 Entering a face feature model; finally, the face recognition result output by the face feature model is obtained by the output unit 56, and can be combined with a face key point other than the face contour, for example, a key point of the nose bone, combined with a plurality of face feature recognition persons. Face type, which improves the reliability of face recognition results.

FIG. 6 is a schematic structural diagram of an intelligent terminal according to an embodiment of the present disclosure. The smart terminal 600 can be any type of smart terminal, such as a mobile phone, a tablet computer, a beauty authentication device, etc., and can perform the operations provided in the embodiments of the present application. Any kind of face face recognition method.

Specifically, referring to FIG. 6, the smart terminal 600 includes:

One or more processors 601 and memory 602, one processor 601 is taken as an example in FIG.

The processor 601 and the memory 602 may be connected by a bus or other means, as exemplified by a bus connection in FIG.

The memory 602 is used as a non-transitory computer readable storage medium, and can be used for storing a non-transitory software program, a non-transitory computer executable program, and a module, such as a program instruction corresponding to the face face recognition method in the embodiment of the present application. / Module (for example, the image acquisition unit 51, the face key extraction unit 52, the reference value determination unit 53, the feature vector construction unit 54, the matching unit 55, the output unit 56, and the model acquisition unit 57) shown in FIG. The processor 601 performs various functional applications and data processing of the device for detecting face defects by running non-transitory software programs, instructions, and modules stored in the memory 602, that is, the face of any of the above method embodiments is implemented. Face recognition method.

The memory 602 may include a storage program area and an storage data area, wherein the storage program area may store an operating system, an application required for at least one function; the storage data area may store data created according to usage of the smart terminal 600, and the like. Moreover, memory 602 can include high speed random access memory, and can also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory 602 can optionally include a processor 601 Remotely set memories that can be connected to the smart terminal 600 via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The one or more modules are stored in the memory 602, and when executed by the one or more processors 601, perform a face face recognition method in any of the above method embodiments, for example, performing the above described diagram The method steps 110 to 160 in 1 and the method steps 151 to 155 in FIG. 3 implement the functions of the units 51-57 in FIG.

The embodiment of the present application further provides a storage medium storing executable instructions executed by one or more processors, for example, by one processor 601 in FIG. 6, which may be The one or more processors perform the face face recognition method in any of the foregoing method embodiments, for example, perform the method steps 110 to 160 in FIG. 1 described above, and the method steps 151 to 155 in FIG. 3 are implemented. The function of units 51-57 in Figure 5.

The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, ie may be located A place, or it can be distributed to multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the embodiment.

Through the description of the above embodiments, those skilled in the art can clearly understand that the various embodiments can be implemented by means of software plus a general hardware platform, and of course, by hardware. A person skilled in the art can understand that all or part of the process of implementing the above embodiments can be completed by a computer program to instruct related hardware, and the program can be stored in a non-transitory computer readable storage medium. The program, when executed, may include the flow of an embodiment of the methods as described above. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

The above products can perform the methods provided by the embodiments of the present application, and have the corresponding functional modules and beneficial effects of the execution method. For technical details that are not described in detail in this embodiment, reference may be made to the method provided by the embodiments of the present application.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not limited thereto; in the idea of the present application, the technical features in the above embodiments or different embodiments may also be combined. The steps may be carried out in any order, and there are many other variations of the various aspects of the present application as described above, which are not provided in the details for the sake of brevity; although the present application has been described in detail with reference to the foregoing embodiments, The skilled person should understand that the technical solutions described in the foregoing embodiments may be modified, or some of the technical features may be equivalently replaced; and the modifications or substitutions do not deviate from the embodiments of the present application. The scope of the technical solution.

Claims

A face face recognition method, comprising:

Obtaining a face image;

Extracting face key points in the face image, wherein the face key points include: a key point of the nose bone, a key point of the mandible, and a key point of the chin;

Determining a reference value of the face image based on the face key point;

Constructing a feature vector of the face image in combination with the face key point and the reference value, wherein the feature vector includes: a face length feature value, a mandible width feature value, and a chin angle feature value, the person A face length feature value is constructed based on the nasal bone key point, the chin key point, and the reference value, the mandible width feature value being constructed based on the mandible key point and the reference value, the chin angle feature value Constructing based on the key points of the mandible and the key points of the chin;

Inputting the feature vector into a face feature model;

Obtaining a face recognition result output by the face feature model.
The facial face recognition method according to claim 1, wherein the face key point further comprises two axillary key points;

Then, determining, according to the face key point, the reference value of the face image, including:

The distance between the two axillary key points is used as a reference value of the face image.
The face face recognition method according to claim 1, wherein the face face recognition method further comprises: before the step of inputting the feature vector into the face feature model:

Obtaining the face feature model.
The facial face recognition method according to claim 2, wherein the acquiring the facial feature model comprises:

Collecting a preset number of face image samples, each of the face image samples being marked with a face mark;

Extracting face key points in the face image sample respectively, wherein the face key points include: Key points of the nasal bone, key points of the mandible, and key points of the chin;

Determining a reference value of each of the face image samples based on the face key points;

Constructing a feature vector of each of the face image samples in combination with the face key point and the reference value, wherein the feature vector includes: a face length feature value, a mandible width feature value, and a chin angle feature value. The face length feature value is constructed based on the nasal bone key point, the chin key point, and the reference value, and the mandible width feature value is constructed based on the mandible key point and the reference value, A chin angle feature value is constructed based on the mandible key point and the chin key point;

The feature vector and the face mark of each of the face image samples are input into a support vector machine model, and the face feature model is trained.
The facial face recognition method according to any one of claims 1 to 4, wherein the face key point further comprises: a sacral key point, the feature vector further comprising: a cheek width feature value, The cheek width feature value is constructed based on the nasal bone key point, the tibia key point, and the reference value.
The facial face recognition method according to claim 5, wherein the feature vector further comprises: a side face length feature value, the side face length feature value is based on the mandible key point, the tibia key The point and the reference value are constructed.
The face face recognition method according to claim 6, wherein the feature vector further comprises: a ratio of the side face length feature value to the face length feature value.
A face face recognition device, comprising:

An image acquisition unit, configured to acquire a face image;

a face key point extracting unit, configured to extract a face key point in the face image, wherein the face key point comprises: a key point of a nose bone, a key point of a mandible, and a key point of a chin;

a reference value determining unit, configured to determine a reference value of the face image based on the face key point;

a feature vector construction unit, configured to construct a feature vector of the face image in combination with the face key point and the reference value, wherein the feature vector includes: a face length feature value, a mandible width feature value, and a chin An angle feature value, the face length feature value is based on the nasal bone key point, the lower a key point of the jaw and a reference value constructed based on the key point of the mandible and the reference value, the chin angle feature value being based on the key point of the mandible and the key point of the chin Construct;

a matching unit, configured to input the feature vector into a face feature model;

And an output unit, configured to acquire a face recognition result output by the face feature model.
The facial face recognition device according to claim 8, wherein the facial face recognition device further comprises:

a model obtaining unit, configured to acquire the face feature model.
The face recognition device according to claim 9, wherein the model acquisition unit comprises:

a face image sample collection module, configured to collect a preset number of face image samples, each of the face image samples being marked with a face mark;

a face key point extraction module, configured to separately extract face key points in the face image sample, wherein the face key points include: a key point of a nose bone, a key point of a mandible, and a key point of a chin;

a reference value determining module, configured to determine a reference value of each of the face image samples based on the face key points respectively;

a feature vector construction module, configured to construct a feature vector of each of the face image samples in combination with the face key point and the reference value, wherein the feature vector comprises: a face length feature value, a mandible width An eigenvalue and a chin angle feature value, the face length feature value being constructed based on the nasal bone key point, the chin key point, and the reference value, the mandible width feature value being based on the mandible key point and Constructing a reference value, the chin angle feature value being constructed based on the mandible key point and the chin key point;

And a training module, configured to input a feature vector and a face mark of each of the face image samples into a support vector machine model, and train the face feature model.
An intelligent terminal, comprising:

At least one processor; and,

a memory communicatively coupled to the at least one processor; wherein

The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method of any of claims 1-7 Face face recognition method.
A storage medium, wherein the storage medium stores executable instructions that, when executed by a smart terminal, cause the smart terminal to perform the human face according to any one of claims 1-7 Face recognition method.