CN113971833A - Multi-angle face recognition method and device, computer main equipment and storage medium - Google Patents
Multi-angle face recognition method and device, computer main equipment and storage medium Download PDFInfo
- Publication number
- CN113971833A CN113971833A CN202111473424.6A CN202111473424A CN113971833A CN 113971833 A CN113971833 A CN 113971833A CN 202111473424 A CN202111473424 A CN 202111473424A CN 113971833 A CN113971833 A CN 113971833A
- Authority
- CN
- China
- Prior art keywords
- face
- angle
- image
- recognition
- attitude
- 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.)
- Pending
Links
Images
Classifications
-
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a multi-angle face recognition method, a multi-angle face recognition device, computer main equipment and a storage medium, wherein the method comprises the following steps: acquiring an image to be recognized, wherein the image to be recognized at least comprises a human face; carrying out angle detection on the face in the image to be recognized to obtain a first attitude angle of the face, wherein the first attitude angle comprises a pitch angle and a yaw angle; according to the first attitude angle of the human face, determining a human face recognition model corresponding to the first attitude angle from a plurality of human face recognition models as a target recognition model; inputting the image to be recognized into the target recognition model to obtain a face recognition result of the image to be recognized; the invention can realize the accurate recognition of the face under each deflection angle, not only improves the recognition accuracy of the face under the deflection angle, but also does not need to collect the portrait of the face during the recognition, thereby improving the convenience of the face recognition.
Description
Technical Field
The invention belongs to the technical field of face recognition, and particularly relates to a multi-angle face recognition method and device, computer main equipment and a storage medium.
Background
With the rapid development of computer equipment, networks and image processing technologies, the traditional eye image recognition mode has been gradually replaced by an image recognition mode automatically performed by a computer, so that the efficiency and accuracy of image recognition are greatly improved, and face recognition performed by intelligent terminals such as computer equipment has been widely applied in various fields, such as mobile phone unlocking, entrance guard unlocking, mobile payment and the like.
The existing face recognition technology mainly comprises the steps of firstly calculating a deflection angle of a face, aligning the face in a roll (representing rotation around a z axis) deflection dimension through affine transformation, and then directly performing face recognition to obtain a recognition result.
Although the existing face recognition technology can correct a deflected face through a face correction algorithm, a part of the face is invisible when the face deflects in the dimensions of yaw (representing rotation around the y axis) and pitch (representing rotation around the x axis), so that the extracted face features are different due to a large difference between information of the corrected face and the original face, and therefore, the existing face recognition technology requires that the face angle and the acquisition angle are consistent, namely, a portrait needs to be acquired by the face, then the face recognition can be carried out, otherwise, the recognition accuracy is affected; therefore, it is urgent to provide a face recognition method supporting multiple angles.
Disclosure of Invention
The invention aims to provide a multi-angle face recognition method, a multi-angle face recognition device, a computer main device and a storage medium, and aims to solve the problem that the existing face recognition technology needs to collect human images on the front face to perform face recognition, otherwise, the recognition accuracy is affected.
In order to achieve the purpose, the invention adopts the following technical scheme:
the invention provides a multi-angle face recognition method, which comprises the following steps:
acquiring an image to be recognized, wherein the image to be recognized at least comprises a human face;
carrying out angle detection on the face in the image to be recognized to obtain a first attitude angle of the face, wherein the first attitude angle comprises a pitch angle and a yaw angle;
according to the first attitude angle of the human face, determining a human face recognition model corresponding to the first attitude angle from a plurality of human face recognition models as a target recognition model;
and inputting the image to be recognized into the target recognition model to obtain a face recognition result of the image to be recognized.
Based on the above disclosure, the invention constructs a plurality of face recognition models, that is, each face recognition model corresponds to a face pose angle range (i.e., deflection angle), and can accurately recognize the face within the pose angle range, so that, when the face recognition is performed, the invention matches the corresponding face recognition model according to the first pose angle by calculating the first pose angle (i.e., deflection angle) of the face, thereby realizing the accurate recognition of the face at each deflection angle, not only improving the recognition accuracy of the face at the deflection angle, but also avoiding acquiring portrait during the recognition of the face, and further improving the convenience of the face recognition.
In one possible design, before acquiring the image to be recognized, the method further includes:
acquiring a training data set, wherein the training data set comprises a plurality of face images;
carrying out angle detection on a plurality of face images in the training data set to obtain a second attitude angle corresponding to each face image in the plurality of face images;
according to a preset angle threshold range and a second attitude angle corresponding to each face image, carrying out sample division on the plurality of face images, and dividing the face images with the second attitude angles in the same angle threshold range into a sample set to obtain a plurality of sample sets;
and allocating a neural network model for each sample set in the plurality of sample sets, and inputting the face image in each sample set into the corresponding neural network model for face recognition training to obtain the plurality of face recognition models.
Based on the disclosure, when performing model training, the method calculates a second attitude angle corresponding to each face image, and then divides a plurality of face images by combining a preset angle threshold range, namely, divides the face images with the second attitude angles in the same angle threshold range together to form a sample set, so that a sample set can be divided for each angle threshold range, and during training, a neural network model is allocated to each sample set for training to obtain a face recognition model corresponding to the angle threshold range; through the design, the face recognition models corresponding to the different angle threshold ranges can be constructed, so that when face recognition is carried out, the recognition models corresponding to the face pose angles are selected for face recognition.
In one possible design, each angle threshold range includes a pitch angle threshold interval and a yaw angle threshold interval, where dividing the face image with the second pose angle within the same angle threshold range into a sample set includes:
and dividing the face image of which the pitch angle in the second attitude angle is within a target pitch angle threshold interval and the face image of which the yaw angle in the second attitude angle is within a target yaw angle threshold interval into a sample set, wherein the target pitch angle threshold interval and the target yaw angle threshold interval are the pitch angle threshold interval and the yaw angle threshold interval within the same angle threshold range.
Based on the above disclosure, when the face image is divided, two angle conditions need to be satisfied simultaneously, so that the face image can be divided into a sample set corresponding to the angle threshold range, that is, the pitch angle in the second attitude angle needs to belong to a pitch angle threshold interval within the same angle threshold range, and the yaw angle needs to belong to a yaw angle threshold interval within the same angle threshold range; therefore, when the human face is identified, only the angle threshold value ranges corresponding to the pitch angle and the yaw angle need to be searched, and the model corresponding to the angle threshold value range is used as the target identification model.
In one possible design, performing angle detection on a face in the image to be recognized to obtain a first pose angle of the face, includes:
performing face key point detection on the image to be recognized to obtain n face key point coordinates, wherein n is a positive integer; and obtaining a first attitude angle of the face according to the n face key point coordinates.
Based on the foregoing disclosure, the present invention discloses a specific calculation method of a first pose angle, that is, a face key point is utilized to perform coordinate transformation, so as to transform the face key point in a camera coordinate system into a face position in a world coordinate system, that is, an euler angle, so that the euler angle is used as the pose angle of the face, so as to select a corresponding recognition model according to the angle in the following.
In one possible design, the n face keypoint coordinates include at least a left eye corner coordinate, a right eye corner coordinate, a nose tip coordinate, a left mouth corner coordinate, a right mouth corner coordinate, and a mandible coordinate.
In one possible design, the pitch angle may have an angular range of [ -180 °, +180 ° ], and the yaw angle may have an angular range of [ -180 °, +180 ° ], wherein positive pitch angle indicates a face-up shift, negative pitch angle indicates a face-down shift, positive yaw angle indicates a face-right yaw, and negative yaw angle indicates a face-left yaw.
In a second aspect, the present invention provides a multi-angle face recognition apparatus, including: the device comprises an acquisition unit, an angle detection unit, a model matching unit and an identification unit;
the acquiring unit is used for acquiring an image to be recognized, wherein the image to be recognized at least comprises a human face;
the angle detection unit is used for carrying out angle detection on the human face in the image to be recognized to obtain a first attitude angle of the human face, wherein the first attitude angle comprises a pitch angle and a yaw angle;
the model matching unit is used for determining a face recognition model corresponding to the first attitude angle from a plurality of face recognition models as a target recognition model according to the first attitude angle of the face;
and the identification unit is used for inputting the image to be identified into the target identification model to obtain a face identification result of the image to be identified.
In one possible design, the apparatus further includes: a sample dividing unit and a model training unit;
the acquisition unit is further configured to acquire a training data set, where the training data set includes a plurality of face images;
the angle detection unit is further configured to perform angle detection on the plurality of face images in the training data set to obtain a second attitude angle corresponding to each of the plurality of face images;
the sample dividing unit is used for carrying out sample division on the plurality of face images according to a preset angle threshold range and a second attitude angle corresponding to each face image, and dividing the face images with the second attitude angles in the same angle threshold range into a sample set to obtain a plurality of sample sets;
the model training unit is used for allocating a neural network model for each sample set in the plurality of sample sets, and inputting the face image in each sample set into the corresponding neural network model for face recognition training to obtain the plurality of face recognition models.
In a third aspect, the present invention provides another multi-angle face recognition apparatus, taking an apparatus as a computer main device as an example, including a memory, a processor and a transceiver, which are sequentially connected in a communication manner, where the memory is used to store a computer program, the transceiver is used to transmit and receive messages, and the processor is used to read the computer program and execute the multi-angle face recognition method as described in the first aspect or any one of the possible designs of the first aspect.
In a fourth aspect, the present invention provides a storage medium having stored thereon instructions for executing the multi-angle face recognition method according to the first aspect or any one of the possible designs of the first aspect when the instructions are run on a computer.
In a fifth aspect, the present invention provides a computer program product comprising instructions which, when run on a computer, cause the computer to perform the multi-angle face recognition method as described in the first aspect or any one of the possible designs in the first aspect.
Drawings
FIG. 1 is a schematic representation of a first attitude angle provided by the present invention;
FIG. 2 is a schematic diagram of a process of training a face recognition model according to the present invention;
FIG. 3 is a schematic view illustrating a process of a multi-angle face recognition method according to the present invention;
FIG. 4 is a schematic structural diagram of a multi-angle face recognition apparatus according to the present invention;
fig. 5 is a schematic structural diagram of a computer main device provided in the present invention.
Detailed Description
The invention is further described with reference to the following figures and specific embodiments. It should be noted that the description of the embodiments is provided to help understanding of the present invention, but the present invention is not limited thereto. Specific structural and functional details disclosed herein are merely illustrative of example embodiments of the invention. This invention may, however, be embodied in many alternate forms and should not be construed as limited to the embodiments set forth herein.
It will be understood that, although the terms first, second, etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element, without departing from the scope of example embodiments of the present invention.
It should be understood that, for the term "and/or" as may appear herein, it is merely an associative relationship that describes an associated object, meaning that three relationships may exist, e.g., a and/or B may mean: a exists alone, B exists alone, and A and B exist at the same time; for the term "/and" as may appear herein, which describes another associative object relationship, it means that two relationships may exist, e.g., a/and B, may mean: a exists independently, and A and B exist independently; in addition, for the character "/" that may appear herein, it generally means that the former and latter associated objects are in an "or" relationship.
Examples
Firstly, a face recognition model based on a face under different attitude angles is constructed for the application, so that when the face recognition is carried out, the face recognition model corresponding to the attitude angle is selected to carry out the face recognition through the attitude angle of the face, the application supports the face recognition under multiple angles, the accuracy rate of the face recognition under multiple angles is improved, a face is not required to be collected at the front face during the recognition, the convenience of the face recognition is improved, and the large-scale popularization and application are facilitated.
Referring to fig. 2, in specific implementation, the method for training a face recognition model in application provided by the present application may include, but is not limited to, the following steps S01 to S04.
S01, a training data set is obtained, wherein the training data set comprises a plurality of face images.
Step S01, acquiring a plurality of face images under different attitude angle states so as to form a training data set, inputting the training data set into a neural network model for training, thereby obtaining a face recognition model so as to perform face recognition on the image to be recognized in the following step; in specific implementation, the following methods can be adopted to acquire a plurality of face images: (1) the method comprises the following steps of collecting face videos of a plurality of people through a camera, wherein the videos contain different attitude angle states of the faces, such as nodding and left-right swinging of the faces; certainly, the face images under different attitude angles can be obtained by processing the collected video frame by frame; (2) the user directly inputs images of a plurality of faces under different attitude angles, so that a training data set is formed.
After the training data set is acquired, the pose angle corresponding to each face image in the training data set may be calculated (in order to facilitate distinguishing from the pose angle of the face in the subsequent image to be recognized, in this embodiment, the pose angle corresponding to each face image is named as a second pose angle), so that the face image is divided according to the angle threshold range where the second pose angle is located, so as to obtain a sample set corresponding to each angle threshold range, so as to obtain a face recognition model in different angle threshold ranges according to the divided sample sets in the subsequent step, where the specific calculation process is as shown in the following step S02.
S02, carrying out angle detection on a plurality of face images in the training data set to obtain a second attitude angle corresponding to each face image in the plurality of face images; in specific implementation, the corresponding second pose angle may be obtained by, but is not limited to, acquiring a face key point in the human body image, and a specific calculation process is as shown in steps S021 and S022 below.
S021, performing face key point detection on each face image to obtain n first face key point coordinates (in order to facilitate distinguishing from key points of a face in a subsequent image to be recognized, the key point coordinates of the face image are named as first face key point coordinates), wherein n is a positive integer; in this embodiment, the face key point detection refers to locating key region positions of the face of a human face, including eyebrows, eyes, a nose, a mouth, a face contour, and the like, given a face image, and in particular, the implementation may use, but is not limited to use: the face key point detection is realized based on an Active Shape Model (ASM) and an Active Appearance Model (AAM), based on Cascaded Shape regression (CPR), or based on a deep learning method, that is, a two-dimensional face mark frame, a two-dimensional face key region, and/or a two-dimensional face key point and the like in a face image are detected.
After n first face key points are obtained, coordinate conversion may be performed, that is, the face position in the camera coordinate system is converted into the face position in the world coordinate system, so as to obtain a second pose angle of the face image, and the specific calculation process is shown in the following step S022.
S022, obtaining a second attitude angle corresponding to each face image according to the n first face key point coordinates; in specific implementation, the following specific steps may be adopted to convert the coordinates of the first face key points into the second pose angles: the first step is as follows: matching the detected coordinates of the first face key points with corresponding face key points in a three-dimensional face model (preset for a user); the second step is that: solving a conversion relation matrix of the original face key points and the corresponding three-dimensional face key points (which can be but is not limited to obtaining a rotation vector by using a solvePnP function of Opencv (which is a cross-platform computer vision and machine learning software library issued based on apache2.0 license), and then obtaining a rotation relation matrix through the rotation vector); the third step: and solving three Euler angles (namely a second attitude angle corresponding to the current face image and comprising a pitch angle pitch, a yaw angle yaw and a roll angle) of the face relative to a camera coordinate system according to the rotation relation matrix.
Referring to FIG. 1, the pitch angle pitch represents the rotation of the face about the x-axis, the yaw angle yaw represents the rotation of the face about the y-axis, and the roll angle roll represents the rotation of the face about the z-axis.
Optionally, for example, the n first-person key point coordinates may include, but are not limited to: the left eye corner coordinate, the right eye corner coordinate, the nose tip coordinate, the left mouth corner coordinate, the right mouth corner coordinate, and the mandible coordinate, of course, the key points may be preset according to the tolerance of the recognition accuracy, and the key points are not limited to the foregoing listed key points.
After the second pose angle of each face in the training data set is obtained, the face image may be divided to obtain a plurality of sample sets divided according to the angle threshold range, as shown in the following step S03.
And S03, carrying out sample division on the plurality of face images according to a preset angle threshold range and a second attitude angle corresponding to each face image, and dividing the face images with the second attitude angles in the same angle threshold range into a sample set to obtain a plurality of sample sets.
As already described above, the second attitude angle includes a pitch angle, a yaw angle and a roll angle, and accordingly, each angle threshold range includes a pitch angle threshold interval and a yaw angle threshold interval.
In specific implementation, the sample set may be divided by a method that divides a face image in which a pitch angle in the second attitude angle is within a target pitch angle threshold interval and divides a face image in which a yaw angle in the second attitude angle is within a target yaw angle threshold interval into one sample set, where the target pitch angle threshold interval and the target yaw angle threshold interval are a pitch angle threshold interval and a yaw angle threshold interval within the same angle threshold range.
The foregoing step S03 and its substeps are described below as a specific example.
Assuming that the angle interval of the pitch angle and the angle interval of the yaw angle are both-180 °, +180 °, and the division of the pitch angle and yaw angle threshold intervals is performed according to 20 ° interval angles, i.e. the number of the angle threshold ranges is 18, the division is performed from-180 °, and the number is 1-18, respectively, so that for the first angle threshold range, the corresponding elevation angle threshold interval is-180 °, -160 °, the corresponding yaw angle threshold interval is-180 °, -160 °, the second angle threshold range is-160 °, -140 °, the corresponding yaw angle threshold interval is-160 °, -140 °, the third angle threshold range is-140 °, -120 °, the corresponding yaw angle threshold interval is-140 °, 120 ° ], a., < 20 °, 0 ° ], an elevation threshold interval corresponding to a ninth angle threshold range, < 20 °, 0 °,20 °, 17 °, 160 °, 140 °, < 140 °, +160 °, 18 °, 160 °, wherein positive pitch angle indicates a human face up-shift (i.e., a human face up), negative pitch angle indicates a human face down-shift (i.e., a human face down), the yaw angle is positive and indicates that the human face deflects rightwards, and the yaw angle is negative and indicates that the human face deflects leftwards.
Meanwhile, assuming that 500 face images are counted in the training data set, wherein the pitch angle in the second attitude angle corresponding to the first face image is 18 ° and the yaw angle is 15 °, the division rule in step S03 is within the tenth angle threshold range, so that the first face image should be divided into sample sets corresponding to the tenth angle threshold range, which may be but is not limited to be named as a tenth sample set, and of course, the division methods of the remaining face images are consistent with the division method of the first face image, which is not described herein.
In this embodiment, in order to distinguish the angle threshold range where the second pose angle corresponding to each face image is located, and perform the division of the sample set, an angle label may be set for each face image, and may be, but is not limited to, the following format: [ { yaw: [0 °,20 ° ] }, pitch: [0 °,20 ° ], flat: face image 1} ], the meaning of the foregoing formula is: the pitch angle of the first face image is in an angle range of [0 degrees and 20 degrees ], and the yaw angle is in an angle range of [0 degrees and 20 degrees ].
Of course, the angle division of the angle threshold range may be specifically set according to actual use, and is not limited to the above-listed examples.
Therefore, through the design, a sample set can be divided for each angle threshold range, so that a neural network model is subsequently allocated for each sample set for training, and thus face recognition models corresponding to different angle threshold ranges are obtained, as shown in the following step S04.
And S04, distributing a neural network model for each sample set in the plurality of sample sets, and inputting the face image in each sample set into the corresponding neural network model for face recognition training to obtain the plurality of face recognition models.
Step S04 is to allocate a neural network model to each sample set, so that the face images in the sample sets are input into the corresponding neural network models for face recognition training, and after the training is completed, the face recognition models corresponding to the angle threshold ranges can be obtained, for example, on the basis of the foregoing example, the face recognition model corresponding to the tenth angle threshold range is used for recognizing the face images with the pitch angle of the face being [0 °,20 ° ] and the yaw angle being [0 °,20 ° ].
In specific implementation, an angle identification tag may be set for each face recognition model, that is, the model is bound to its corresponding angle threshold range, for example, a face recognition model a — a first angle threshold range (that is, the model a is used for recognizing a face image with an elevation angle threshold range of [ -180 °, -160 ° ], and a yaw angle threshold range of [ -180 °, -160 ° ]); therefore, when in recognition, the models can be quickly matched according to the attitude angle of the human face, and certainly, the serial numbers of the subsequent human face recognition models are analogized by the same way and named by 26 letters.
In addition, example Neural network models may use, but are not limited to, Networks such as a Yolo (a Single Neural network based object detection system proposed by Joseph Redmon and AliFarhadi et al in 2015), an SSD (solid State disk Multi Box Detector) network, or an R-CNN (Convolutional Neural network) network.
Therefore, the plurality of face recognition models constructed through the steps S01-S04 and the sub-steps are only required to be matched through the attitude angles during recognition, and through the design, the face recognition method not only supports the face recognition under multiple angles, but also does not need to collect human images on the front face during recognition, so that the convenience of the face recognition is improved.
Referring to fig. 3, in the second aspect of the present embodiment, on the face recognition model constructed in the first aspect of the present embodiment, a face is recognized on an image to be recognized, and the recognition process is as shown in the following steps S1 to S4.
S1, obtaining an image to be recognized, wherein the image to be recognized at least comprises one face.
S2, carrying out angle detection on the face in the image to be recognized to obtain a first attitude angle of the face, wherein the first attitude angle comprises a pitch angle and a yaw angle.
In step S2, the process of calculating the first pose angle of the face is the same as the process of calculating the second pose angle, and the specific process can be shown in step S021 and step S022, which are not repeated herein.
And S3, according to the first attitude angle of the human face, determining a human face recognition model corresponding to the first attitude angle from a plurality of human face recognition models, and using the human face recognition model as a target recognition model.
Step S3 is to determine an angle threshold range to which the first attitude angle belongs, for example, based on the foregoing example, if the pitch angle in the first attitude angle is 18 ° and the yaw angle is 15 °, the first attitude angle belongs to a tenth angle threshold range, and the corresponding face recognition model is a face recognition model J; therefore, the face recognition model J can be used to perform face recognition on the image to be recognized, so as to obtain a face recognition result, as shown in the following step S4.
And S4, inputting the image to be recognized into the target recognition model to obtain a face recognition result of the image to be recognized.
Therefore, by the multi-angle face recognition method described in detail in the steps S1-S4, the method supports the face recognition under multiple angles, not only improves the accuracy of the face recognition under multiple angles, but also does not need to collect the portrait on the front face during the recognition, thereby improving the convenience of the face recognition.
As shown in fig. 4, a third aspect of the present embodiment provides a hardware device for implementing the multi-angle face recognition method in the first aspect or the second aspect of the embodiments, including: the device comprises an acquisition unit, an angle detection unit, a model matching unit and a recognition unit.
The acquiring unit is used for acquiring an image to be recognized, wherein the image to be recognized at least comprises a human face.
The angle detection unit is used for carrying out angle detection on the human face in the image to be recognized to obtain a first attitude angle of the human face, wherein the first attitude angle comprises a pitch angle and a yaw angle.
And the model matching unit is used for determining a face recognition model corresponding to the first attitude angle from a plurality of face recognition models as a target recognition model according to the first attitude angle of the face.
And the identification unit is used for inputting the image to be identified into the target identification model to obtain a face identification result of the image to be identified.
In one possible design, the apparatus further includes: the device comprises a sample dividing unit and a model training unit.
The acquiring unit is further configured to acquire a training data set, where the training data set includes a plurality of face images.
The angle detection unit is further configured to perform angle detection on the multiple face images in the training data set to obtain a second attitude angle corresponding to each of the multiple face images.
The sample division unit is used for carrying out sample division on the plurality of face images according to a preset angle threshold range and a second attitude angle corresponding to each face image, and dividing the face images with the second attitude angles in the same angle threshold range into a sample set to obtain a plurality of sample sets.
The model training unit is used for allocating a neural network model for each sample set in the plurality of sample sets, and inputting the face image in each sample set into the corresponding neural network model for face recognition training to obtain the plurality of face recognition models.
For the working process, the working details, and the technical effects of the hardware apparatus provided in this embodiment, reference may be made to the first aspect or the second aspect of the embodiment, which is not described herein again.
As shown in fig. 5, a fourth aspect of this embodiment provides another multi-angle face recognition apparatus, taking the apparatus as a computer main device as an example, including: the face recognition system comprises a memory, a processor and a transceiver which are sequentially connected in a communication manner, wherein the memory is used for storing a computer program, the transceiver is used for transceiving messages, and the processor is used for reading the computer program and executing the multi-angle face recognition method according to the first aspect or the second aspect of the embodiment.
For example, the Memory may include, but is not limited to, a Random Access Memory (RAM), a Read Only Memory (ROM), a Flash Memory (Flash Memory), a First In First Out (FIFO), and/or a First In Last Out (FILO), and the like; in particular, the processor may include one or more processing cores, such as a 4-core processor, an 8-core processor, and so on. The processor may be implemented in at least one hardware form of a DSP (Digital Signal Processing), an FPGA (Field-Programmable Gate Array), and a PLA (Programmable Logic Array).
Meanwhile, the processor may also include a main processor and a coprocessor, where the main processor is a processor for Processing data in an awake state, and is also called a Central Processing Unit (CPU); a coprocessor is a low power processor for processing data in a standby state. In some embodiments, the processor may be integrated with a GPU (Graphics Processing Unit) which is responsible for rendering and drawing contents required to be displayed on the display screen, for example, the processor may not be limited to a processor adopting a model STM32F105 series microprocessor, a reduced instruction set computer (RSIC) microprocessor, an architecture processor such as X86, or a processor integrated with an embedded neural Network Processor (NPU); the transceiver may be, but is not limited to, a wireless fidelity (WIFI) wireless transceiver, a bluetooth wireless transceiver, a General Packet Radio Service (GPRS) wireless transceiver, a ZigBee wireless transceiver (ieee802.15.4 standard-based low power local area network protocol), a 3G transceiver, a 4G transceiver, and/or a 5G transceiver, etc. In addition, the device may also include, but is not limited to, a power module, a display screen, and other necessary components.
For the working process, the working details, and the technical effects of the computer main device provided in this embodiment, reference may be made to the first aspect or the second aspect of the embodiment, which is not described herein again.
A fifth aspect of the present embodiment provides a storage medium storing instructions including the multi-angle face recognition method according to the first aspect or the second aspect, that is, the storage medium stores instructions that, when executed on a computer, perform the multi-angle face recognition method according to the first aspect or the second aspect.
The storage medium refers to a carrier for storing data, and may include, but is not limited to, a floppy disk, an optical disk, a hard disk, a flash Memory, a flash disk and/or a Memory Stick (Memory Stick), etc., and the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.
For the working process, the working details, and the technical effects of the storage medium provided in this embodiment, reference may be made to the first aspect or the second aspect of the embodiment, which is not described herein again.
A sixth aspect of the present embodiments provides a computer program product comprising instructions for causing a computer to perform the method for multi-angle face recognition according to the first or second aspect of the embodiments when the instructions are run on the computer, wherein the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable apparatus.
Finally, it should be noted that: the above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A multi-angle face recognition method is characterized by comprising the following steps:
acquiring an image to be recognized, wherein the image to be recognized at least comprises a human face;
carrying out angle detection on the face in the image to be recognized to obtain a first attitude angle of the face, wherein the first attitude angle comprises a pitch angle and a yaw angle;
according to the first attitude angle of the human face, determining a human face recognition model corresponding to the first attitude angle from a plurality of human face recognition models as a target recognition model;
and inputting the image to be recognized into the target recognition model to obtain a face recognition result of the image to be recognized.
2. The method of claim 1, wherein prior to acquiring the image to be identified, the method further comprises:
acquiring a training data set, wherein the training data set comprises a plurality of face images;
carrying out angle detection on a plurality of face images in the training data set to obtain a second attitude angle corresponding to each face image in the plurality of face images;
according to a preset angle threshold range and a second attitude angle corresponding to each face image, carrying out sample division on the plurality of face images, and dividing the face images with the second attitude angles in the same angle threshold range into a sample set to obtain a plurality of sample sets;
and allocating a neural network model for each sample set in the plurality of sample sets, and inputting the face image in each sample set into the corresponding neural network model for face recognition training to obtain the plurality of face recognition models.
3. The method of claim 2, wherein each angular threshold range comprises a pitch angle threshold interval and a yaw angle threshold interval, and wherein dividing the face image having the second pose angle within the same angular threshold range into a sample set comprises:
and dividing the face image of which the pitch angle in the second attitude angle is within a target pitch angle threshold interval and the face image of which the yaw angle in the second attitude angle is within a target yaw angle threshold interval into a sample set, wherein the target pitch angle threshold interval and the target yaw angle threshold interval are the pitch angle threshold interval and the yaw angle threshold interval within the same angle threshold range.
4. The method of claim 1, wherein performing angle detection on a face in the image to be recognized to obtain a first pose angle of the face comprises:
performing face key point detection on the image to be recognized to obtain n face key point coordinates, wherein n is a positive integer;
and obtaining a first attitude angle of the face according to the n face key point coordinates.
5. The method of claim 4, wherein the n face keypoint coordinates comprise at least a left eye corner coordinate, a right eye corner coordinate, a nose tip coordinate, a left mouth corner coordinate, a right mouth corner coordinate, and a mandible coordinate.
6. The method of claim 1, wherein the pitch angle has an angular interval of [ -180 °, +180 ° ], and the yaw angle has an angular interval of [ -180 °, +180 ° ], wherein positive pitch angle indicates a face-up shift, negative pitch angle indicates a face-down shift, positive yaw angle indicates a face-right yaw, and negative yaw angle indicates a face-left yaw.
7. A multi-angle face recognition device, comprising: the device comprises an acquisition unit, an angle detection unit, a model matching unit and an identification unit;
the acquiring unit is used for acquiring an image to be recognized, wherein the image to be recognized at least comprises a human face;
the angle detection unit is used for carrying out angle detection on the human face in the image to be recognized to obtain a first attitude angle of the human face, wherein the first attitude angle comprises a pitch angle and a yaw angle;
the model matching unit is used for determining a face recognition model corresponding to the first attitude angle from a plurality of face recognition models as a target recognition model according to the first attitude angle of the face;
and the identification unit is used for inputting the image to be identified into the target identification model to obtain a face identification result of the image to be identified.
8. The apparatus of claim 7, wherein the apparatus further comprises: a sample dividing unit and a model training unit;
the acquisition unit is further configured to acquire a training data set, where the training data set includes a plurality of face images;
the angle detection unit is further configured to perform angle detection on the plurality of face images in the training data set to obtain a second attitude angle corresponding to each of the plurality of face images;
the sample dividing unit is used for carrying out sample division on the plurality of face images according to a preset angle threshold range and a second attitude angle corresponding to each face image, and dividing the face images with the second attitude angles in the same angle threshold range into a sample set to obtain a plurality of sample sets;
the model training unit is used for allocating a neural network model for each sample set in the plurality of sample sets, and inputting the face image in each sample set into the corresponding neural network model for face recognition training to obtain the plurality of face recognition models.
9. A computer master device, comprising: the face recognition system comprises a memory, a processor and a transceiver which are sequentially connected in a communication manner, wherein the memory is used for storing a computer program, the transceiver is used for receiving and sending messages, and the processor is used for reading the computer program and executing the multi-angle face recognition method according to any one of claims 1 to 6.
10. A storage medium having stored thereon instructions for performing the method of face recognition from multiple angles according to any one of claims 1 to 6 when the instructions are run on a computer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111473424.6A CN113971833A (en) | 2021-11-29 | 2021-11-29 | Multi-angle face recognition method and device, computer main equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111473424.6A CN113971833A (en) | 2021-11-29 | 2021-11-29 | Multi-angle face recognition method and device, computer main equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113971833A true CN113971833A (en) | 2022-01-25 |
Family
ID=79590664
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111473424.6A Pending CN113971833A (en) | 2021-11-29 | 2021-11-29 | Multi-angle face recognition method and device, computer main equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113971833A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114241590A (en) * | 2022-02-28 | 2022-03-25 | 深圳前海清正科技有限公司 | Self-learning face recognition terminal |
| CN115359589A (en) * | 2022-08-08 | 2022-11-18 | 珠海格力电器股份有限公司 | Control method and device of intelligent door lock, electronic device and storage medium |
| CN116071836A (en) * | 2023-03-09 | 2023-05-05 | 山东科技大学 | Deep learning-based crewman abnormal behavior detection and identity recognition method |
| WO2023231400A1 (en) * | 2022-05-31 | 2023-12-07 | 青岛云天励飞科技有限公司 | Method and apparatus for predicting facial angle, and device and readable storage medium |
-
2021
- 2021-11-29 CN CN202111473424.6A patent/CN113971833A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114241590A (en) * | 2022-02-28 | 2022-03-25 | 深圳前海清正科技有限公司 | Self-learning face recognition terminal |
| CN114241590B (en) * | 2022-02-28 | 2022-07-22 | 深圳前海清正科技有限公司 | Self-learning face recognition terminal |
| WO2023231400A1 (en) * | 2022-05-31 | 2023-12-07 | 青岛云天励飞科技有限公司 | Method and apparatus for predicting facial angle, and device and readable storage medium |
| CN115359589A (en) * | 2022-08-08 | 2022-11-18 | 珠海格力电器股份有限公司 | Control method and device of intelligent door lock, electronic device and storage medium |
| CN115359589B (en) * | 2022-08-08 | 2023-10-10 | 珠海格力电器股份有限公司 | Control method and device of intelligent door lock, electronic device and storage medium |
| CN116071836A (en) * | 2023-03-09 | 2023-05-05 | 山东科技大学 | Deep learning-based crewman abnormal behavior detection and identity recognition method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113971833A (en) | Multi-angle face recognition method and device, computer main equipment and storage medium | |
| EP3407294B1 (en) | Information processing method, device, and terminal | |
| CN107220995B (en) | Improved method of ICP (inductively coupled plasma) rapid point cloud registration algorithm based on ORB (object-oriented bounding Box) image features | |
| CN106651767A (en) | Panoramic image obtaining method and apparatus | |
| CN110852182B (en) | Depth video human body behavior recognition method based on three-dimensional space time sequence modeling | |
| CN107748890A (en) | A kind of visual grasping method, apparatus and its readable storage medium storing program for executing based on depth image | |
| CN105069809B (en) | A kind of camera localization method and system based on planar hybrid marker | |
| CN111191582B (en) | Three-dimensional target detection method, detection device, terminal device and computer readable storage medium | |
| CN109948397A (en) | A kind of face image correcting method, system and terminal device | |
| CN109190516A (en) | A kind of static gesture identification method based on volar edge contour vectorization | |
| CN103971087B (en) | Method and device for searching and recognizing traffic signs in real time | |
| CN111612841A (en) | Target positioning method and device, mobile robot and readable storage medium | |
| CN110163087A (en) | A kind of human face posture recognition methods and system | |
| CN102222341A (en) | Method and device for detecting motion characteristic point and method and device for detecting motion target | |
| CN110363817A (en) | Object pose estimation method, electronic equipment and medium | |
| CN103714530A (en) | Vanishing point detection and image correction method | |
| CN110163831A (en) | The object Dynamic Display method, apparatus and terminal device of three-dimensional sand table | |
| CN110361005A (en) | Positioning method, positioning device, readable storage medium and electronic equipment | |
| CN105893981A (en) | Human face posture correction method | |
| CN110222651A (en) | A kind of human face posture detection method, device, terminal device and readable storage medium storing program for executing | |
| CN114049674A (en) | Three-dimensional face reconstruction method, device and storage medium | |
| CN108052869B (en) | Lane line recognition method, lane line recognition device and computer-readable storage medium | |
| CN108647640A (en) | The method and electronic equipment of recognition of face | |
| CN111046796A (en) | Low-cost space gesture control method and system based on double-camera depth information | |
| CN106022236A (en) | Action identification method based on human body contour |
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 |