WO2020164284A1 - 基于平面检测的活体识别方法、装置、终端及存储介质 - Google Patents

基于平面检测的活体识别方法、装置、终端及存储介质 Download PDF

Info

Publication number
WO2020164284A1
WO2020164284A1 PCT/CN2019/118553 CN2019118553W WO2020164284A1 WO 2020164284 A1 WO2020164284 A1 WO 2020164284A1 CN 2019118553 W CN2019118553 W CN 2019118553W WO 2020164284 A1 WO2020164284 A1 WO 2020164284A1
Authority
WO
WIPO (PCT)
Prior art keywords
posture
local
feature points
living body
plane
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.)
Ceased
Application number
PCT/CN2019/118553
Other languages
English (en)
French (fr)
Inventor
王路生
陆进
陈斌
宋晨
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Ping An Technology Shenzhen Co Ltd
Original Assignee
Ping An Technology Shenzhen Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Ping An Technology Shenzhen Co Ltd filed Critical Ping An Technology Shenzhen Co Ltd
Publication of WO2020164284A1 publication Critical patent/WO2020164284A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/161—Detection; Localisation; Normalisation
    • G06V40/166—Detection; Localisation; Normalisation using acquisition arrangements
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/168—Feature extraction; Face representation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/40—Spoof detection, e.g. liveness detection
    • G06V40/45—Detection of the body part being alive

Definitions

  • This application belongs to the technical field of living body identification, and in particular relates to a living body identification method, device, terminal and storage medium based on plane detection.
  • living body recognition is mainly used to determine the true physiological characteristics of the detected object in some identity verification scenarios.
  • a face recognition application the combined actions of blinking, opening the mouth, shaking the head, and nodding are detected to verify whether the current detection object is a real living person. It can effectively resist common live attacks such as photos, face changes, masks, occlusions, and screen remakes, so as to identify fraudulent behaviors and protect the interests of users.
  • the existing face live detection methods mainly have the following problems: first, the calculation time is expensive, for example, it needs to use the three-dimensional depth information and use the optical flow method to calculate the change of the non-rigid operation of the face, and the calculation process is complicated; second, the need
  • additional biometric identification devices for cooperative identification requires high equipment costs. For example, additional infrared human body detection devices need to be used to detect the temperature of the detected object, or sound collection devices need to be used for joint voice recognition. It can be seen that the living body identification method in the prior art has the problem of complicated calculation process or high cost of additional identification equipment.
  • the present application provides a living body identification method, device, terminal and storage medium based on plane detection to solve the complicated calculation process of the living body identification method in the prior art or the need to equip additional identification equipment with high cost problem.
  • the first aspect of the embodiments of the present application provides a living body recognition method based on plane detection, including:
  • This application obtains the plane image of the detection object, uses the facial feature points on the plane image to determine the overall head posture and the partial head posture of the detection object on the plane image, and determines the overall head posture and the partial head posture The posture difference is used to determine whether the detection object is a living body.
  • This application can solve the problem of complicated calculation process in the existing living body recognition method or high cost of additional recognition equipment; that is, on the one hand, the recognition process of this application may not Involving three-dimensional depth information, the calculation process is simplified, which is beneficial to improve the efficiency of living body identification; on the other hand, the identification process of the present application does not need to introduce additional biometric identification equipment, which reduces the cost of equipment for living body identification.
  • FIG. 1 is a flowchart of an embodiment of a living body recognition method based on plane detection in an embodiment of the application;
  • FIG. 2 is a flowchart of an embodiment of step S103 in the embodiment shown in FIG. 1;
  • FIG. 3 is a flowchart of another embodiment of a living body recognition method based on plane detection in an embodiment of the application;
  • FIG. 4 is a structural diagram of an embodiment of a living body identification device based on plane detection in an embodiment of the application;
  • FIG. 5 is a schematic block diagram of an identification terminal in an embodiment of the application.
  • FIG. 1 is a flowchart of an embodiment of a living body recognition method based on plane detection in an embodiment of this application, which may include:
  • step S101 a plane image of the detection target is acquired.
  • a planar image of the detection object is first acquired.
  • the planar image refers to a two-dimensional image, which can be specifically acquired through a terminal equipped with an image sensor.
  • a mobile phone terminal equipped with a camera can obtain a plane image of the detection object.
  • step S102 the facial feature points on the plane image are extracted.
  • the facial feature points refer to pixels or sets of pixel points that can be used to embody facial features on a flat image.
  • Each facial feature point can reflect a feature of a human face; specifically, it is reflected in In a planar image, each facial feature point can be a pixel, or a collection of multiple adjacent pixels, for example, a pixel block formed by multiple adjacent pixels.
  • the extracted facial feature points may be multiple predefined facial feature points.
  • the extracted facial feature points may include the tip of the nose, the chin, the left corner of the left eye, the right corner of the right eye, and the mouth.
  • the left corner of the face and the right corner of the mouth are the corresponding pixels or sets of pixels on the plane image.
  • step S103 the overall head posture corresponding to the planar image is determined based on the extracted facial feature points.
  • the head posture of the person on it can include pitch (up and down, for example, head down and up), yaw (left and right, for example, left or left head) Part), roll (rotation in the plane, such as bowing to the left or right) three angles.
  • the overall head posture is the orientation of the face in the three-dimensional space, and the orientation can be one of the above three angles in the three-dimensional space.
  • the different orientations of the human face in the three-dimensional space are reflected on the plane image, which is mainly reflected in the different position distribution of the feature points of the human face. Therefore, the orientation of the face on the plane image in the three-dimensional space can be determined according to the position distribution of the extracted facial feature points, that is, according to the position distribution of the face feature points on the plane image, the corresponding position of the plane image can be obtained.
  • the overall head posture in three-dimensional space is mainly reflected in the different position distribution of the feature points of the human face.
  • a neural network for overall head posture estimation can be established, and a plane image of the determined head posture can be used as a training sample to train and deep-learn the established neural network, and the head to be determined
  • the plane image of the posture is input into the trained neural network to determine the overall head posture of the plane image.
  • the appearance-based method can also be used to estimate the overall head posture by using the apparent features of the face on the plane image.
  • an embodiment of the foregoing step S103 may include:
  • Step S1031 Obtain the position distribution of the extracted facial feature points on the plane image to obtain a first position distribution.
  • Step S1032 Adjust the preset posture of the three-dimensional face model, and obtain the projection position distribution of the facial feature points on the three-dimensional face model on the two-dimensional plane during the adjustment process to obtain a second position distribution.
  • the facial feature points on the three-dimensional face model correspond to the facial feature points on the extracted planar image one-to-one.
  • Step S1033 Obtain the spatial pose of the three-dimensional face model when the second position distribution is consistent with the first position distribution, to obtain the target pose.
  • Step S1031 Determine the target pose as the overall head pose corresponding to the planar image.
  • a standard three-dimensional face model can be set in advance, and the three-dimensional face model can be rotated from a reference direction (for example, the forward direction) to adjust its face orientation, and the three-dimensional face model can be monitored.
  • the position change of the facial feature points is obtained when the projection position distribution of the facial feature points on the three-dimensional face model on the two-dimensional plane is the same or nearly the same as the position distribution of the facial feature points on the extracted plane image.
  • the azimuth angle rotation information of the three-dimensional face model, the azimuth angle rotation information at this time is the head posture of the three-dimensional face model, that is, the overall head posture corresponding to the planar image.
  • the projection calculation of the facial feature points on the three-dimensional face model on the two-dimensional plane can use the camera calibration method and the direct linear transformation (Direct Linear Transform).
  • Transform, DLT Direct Linear Transform
  • DLT Direct Linear Transform
  • step S104 the extracted face feature points are divided into multiple local feature groups, and the local head pose corresponding to each local feature group is determined.
  • the extracted face feature points are grouped to obtain multiple local feature groups. For each local feature group, by adjusting the three-dimensional face model in the same manner as described above, the corresponding local feature group can be obtained Local head posture.
  • step S104 may include:
  • the local spatial posture corresponding to each local feature group when the three-dimensional face model is in the target posture is determined as the local head posture corresponding to the local feature group.
  • the specified number is an integer not less than 3, and the specified number is less than the number of extracted facial feature points.
  • the specified number may be 3, that is, every three facial feature points As a local feature group.
  • three facial feature points, the tip of the nose, the left corner of the left eye, and the right corner of the right eye are divided into a group to obtain a first local feature group.
  • the three-dimensional face model is in the above-mentioned target pose
  • the three-dimensional face The orientation of the plane corresponding to the three face feature points of the nose tip, the left corner of the left eye, and the right corner of the right eye on the model is determined as the local head pose corresponding to the first local feature group.
  • the three facial feature points of the chin, the left corner of the left eye, and the right corner of the right eye are divided into a group to obtain a second local feature group.
  • the three-dimensional face model is in the above-mentioned target pose
  • the three-dimensional face model The orientation of the plane where the three face feature points of the chin, the left corner of the left eye, and the right corner of the right eye are located is determined as the local head pose corresponding to the second local feature group.
  • step S105 the posture difference between the local head posture and the overall head posture is calculated.
  • the posture difference between the local head posture corresponding to each local feature group and the overall head posture of the planar image is counted, and it is further possible to determine whether the detection object is a living body through the posture difference calculated.
  • step S105 may include:
  • the overall head posture and the partial head posture are vectorized, and the overall head posture is represented by the posture vector when the three-dimensional face model is in the aforementioned target posture, and the three-dimensional face model is in the aforementioned target posture.
  • the plane normal vector of the plane where each local feature group is located is used to represent the local head pose, and the angle between each plane normal vector and the pose vector when the three-dimensional face model is in the above target pose is calculated, and the local head is represented by the included angle
  • the posture difference between the posture and the overall head posture The larger the angle, the greater the posture difference between the local head posture and the overall head posture. The smaller the angle, the larger the posture between the local head posture and the overall head posture. The smaller the difference.
  • step S106 it is determined whether the detection object is a living body based on the posture difference.
  • the posture difference reflects the difference between the local plane of the head and the overall posture. In fact, for a living object, the posture difference usually does not exceed a threshold. Then, the posture difference is compared with the threshold. By comparison, it can be judged whether the detection object is a living body. For example, when the posture difference between the local head posture corresponding to a certain local feature group and the overall head posture is greater than a set threshold, it can be judged whether the detected correspondence is a living body.
  • step 106 may include:
  • the detection object is not a living body
  • the detection object is a living body.
  • the first number of local feature groups whose angle with the pose vector is greater than the first preset threshold can be counted.
  • the first number is greater than the first specified value, It is determined that the detection object this time is not a living body, and when the first number is not greater than the first specified value, it is determined that the detection object this time is a living body.
  • the head posture of the face follows the plane of the display and the camera (which collects the plane image).
  • the change of the angle between the device) will basically remain unchanged. This is because the video (image) of the attacking avatar is guaranteed.
  • the angle between the local plane (partial head posture) and the camera pointing will be Incorrect, that is, there will be an angle difference between the two vectors (posture difference).
  • the embodiment of the present application calculates the angle between the multiple sets of plane normal vectors (local head posture) and the overall head posture vector (overall head posture), and determines the living body based on the size of the included angle.
  • the angle between each group of plane normal vectors corresponding to the living target and the overall head posture vector will be kept within a preset threshold, so the living body can be judged based on the size of the angle.
  • this application obtains a planar image of the detection object, uses the facial feature points on the planar image to determine the overall head posture and the partial head posture of the detection object on the planar image, and determines the overall head posture
  • the posture difference of the local head posture is used to determine whether the detection object is a living body.
  • the present application can solve the problem of complicated calculation process in the existing living body recognition method or high cost of additional recognition equipment; that is, on the one hand, the present application
  • the recognition process does not involve three-dimensional depth information, which simplifies the calculation process and helps improve the efficiency of living body recognition.
  • the recognition process of the present application does not need to introduce additional biometric recognition equipment, which reduces the cost of equipment for living body recognition.
  • FIG. 3 is another embodiment of a living body recognition method based on plane detection in an embodiment of this application, which may include:
  • Step S301 Obtain a plane image of the detection object.
  • Step S302 Extract the facial feature points on the plane image.
  • Step S303 Determine the overall head posture corresponding to the planar image based on the extracted facial feature points.
  • Step S304 Divide the extracted facial feature points into multiple local feature groups, and determine the local head pose corresponding to each local feature group.
  • Step S305 Calculate the posture difference between the local head posture and the overall head posture.
  • steps S301 to S305 can be specifically referred to steps S101 to S105 in the embodiment shown in FIG. 1, and details are not described herein again.
  • Step S306 Output an action instruction for instructing the detection object to perform a specified head movement.
  • Step S307 Monitoring the change rate of the posture difference of the detection object when performing the head movement
  • Step S308 Determine whether the detection object is a living body based on the change rate of the posture difference.
  • the detection object in order to further improve the accuracy of recognition, in actual implementation, the detection object may also be required to perform one or more specified posture changes, such as left, right, up and down.
  • one or more specified posture changes such as left, right, up and down.
  • the attack image cannot be fixed, this can improve the algorithm's performance. Effectiveness.
  • the angle change rate of each plane normal vector and the head posture vector can also be evaluated in real time.
  • the angle change rate reflects the detection object's head movement.
  • the change rate of the posture difference between the overall head posture and the local head posture can also be judged in vivo according to the angle change rate.
  • a plane image after the detection object performs the head movement can be obtained, and a posture difference can be obtained based on the plane image in the same manner as above, and the posture difference is compared with that before the head movement.
  • the attitude difference corresponding to the plane image is compared, and the attitude difference change rate can be obtained.
  • step S308 may include:
  • the second number is greater than a second specified value, it is determined that the detection object is not a living body
  • the detection object is a living body.
  • the second number of local feature groups whose posture difference change rate is greater than the second preset threshold is counted.
  • the second number is greater than the second specified value, it is determined that the detection object is not a living body this time.
  • the second number is not greater than the second specified value, it is determined that the object of detection this time is a living body.
  • this application obtains a planar image of the detection object, uses the facial feature points on the planar image to determine the overall head posture and the partial head posture of the detection object on the planar image, and determines the overall head posture
  • the posture difference of the local head posture is used to determine whether the detection object is a living body.
  • the present application can solve the problem of complicated calculation process in the existing living body recognition method or high cost of additional recognition equipment; that is, on the one hand, the present application
  • the recognition process does not involve three-dimensional depth information, which simplifies the calculation process and helps improve the efficiency of living body recognition.
  • the recognition process of the present application does not need to introduce additional biometric recognition equipment, which reduces the cost of equipment for living body recognition.
  • FIG. 4 shows a structural diagram of an embodiment of a living body identification device based on plane detection provided in an embodiment of the present application.
  • the living body recognition device 4 based on plane detection may include: a plane image acquisition unit 41, a feature point extraction unit 42, an overall posture determination unit 43, a local posture determination unit 44, a posture difference calculation unit 45, and a living body judgment unit 46 .
  • the plane image acquiring unit 41 is used to acquire a plane image of the face of the detection object
  • the feature point extraction unit 42 is configured to extract the face feature points of the face plane image
  • the overall posture determination unit 43 is configured to determine the overall head posture corresponding to the planar image based on the extracted facial feature points;
  • the local posture determination unit 44 is configured to divide the extracted facial feature points into multiple local feature groups, and determine the local head posture corresponding to each local feature group;
  • a posture difference calculation unit 45 configured to calculate the posture difference between the local head posture and the overall head posture
  • the living body judging unit 46 is configured to judge whether the detection object is a living body based on the posture difference.
  • the living body identification device 4 based on plane detection may further include:
  • the first acquiring unit is configured to acquire the position distribution of the extracted facial feature points on the plane image to obtain the first position distribution
  • the second acquisition unit is used to adjust the pose of the preset three-dimensional face model, and during the adjustment process, acquire the position distribution of the projection of the facial feature points on the three-dimensional face model on the two-dimensional plane to obtain the second Location distribution, wherein the facial feature points on the three-dimensional face model correspond to the facial feature points on the extracted plane image one-to-one;
  • a third acquiring unit configured to acquire the spatial pose of the three-dimensional face model when the second position distribution is consistent with the first position distribution, to obtain a target pose
  • the overall posture determining unit 43 is specifically configured to determine the target posture as the overall head posture corresponding to the planar image.
  • the living body identification device 4 based on plane detection may further include:
  • the feature group dividing unit is configured to divide the extracted face feature points into a plurality of local feature groups by using a specified number of face feature points as a group, wherein the specified number is less than the number of the extracted face feature points ;
  • a spatial pose calculation unit configured to calculate a local spatial pose corresponding to each local feature group when the three-dimensional face model is in the target pose
  • the local posture determination unit 44 is specifically configured to determine the local spatial posture corresponding to each local feature group when the three-dimensional face model is in the target posture as the local head posture corresponding to the local feature group.
  • the living body identification device 4 based on plane detection may further include:
  • a first vector acquiring unit configured to acquire a pose vector when the three-dimensional face model is in the target pose
  • the second vector acquisition unit is configured to acquire the plane normal vector of each local feature group when the three-dimensional face model is in the target pose;
  • the posture difference calculation unit 45 is specifically configured to calculate the angle between each plane normal vector and the posture vector, wherein the size of the angle represents the magnitude of the posture difference.
  • the living body identification device 4 based on plane detection may further include:
  • the first quantity statistics unit is configured to count the first quantity of the local feature group whose angle with the pose vector is greater than the first preset threshold
  • the living body judging unit 46 is specifically configured to, if the first number is greater than a first specified value, judge that the detection object is not a living body; if the first number is not greater than the first specified value, judge that the detection The object is a living body.
  • the living body identification device 4 based on plane detection may further include:
  • An action instruction unit for outputting an action instruction for instructing the detection object to perform a designated head movement
  • a difference monitoring unit configured to monitor the change rate of the posture difference when the detection object performs the head movement
  • the living body judging unit 46 is further configured to judge whether the detection object is a living body based on the change rate of the posture difference.
  • the living body identification device 4 based on plane detection may further include:
  • the second quantity statistics unit is configured to count the second quantity of the local feature groups whose posture difference change rate is greater than a second preset threshold
  • the living body judging unit 46 is also specifically configured to, if the second number is greater than a second specified value, judge that the detection object is not a living body; if the second number is not greater than the second specified value, judge the The detection object is a living body.
  • this application obtains a planar image of the detection object, uses the facial feature points on the planar image to determine the overall head posture and the partial head posture of the detection object on the planar image, and determines the overall head posture
  • the posture difference of the local head posture is used to determine whether the detection object is a living body.
  • the present application can solve the problem of complicated calculation process in the existing living body recognition method or high cost of additional recognition equipment; that is, on the one hand, the present application
  • the recognition process does not involve three-dimensional depth information, which simplifies the calculation process and helps improve the efficiency of living body recognition.
  • the recognition process of the present application does not need to introduce additional biometric recognition equipment, which reduces the cost of equipment for living body recognition.
  • FIG. 5 shows a schematic block diagram of an identification terminal provided by an embodiment of the present application. For ease of description, only parts related to the embodiment of the present application are shown.
  • the identification terminal 5 may be a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud server.
  • the identification terminal 5 may include: a processor 50, a memory 51, and computer-readable instructions 52 stored in the memory 51 and running on the processor 50, such as those that execute the above-mentioned living body identification method based on plane detection.
  • Computer readable instructions When the processor 50 executes the computer-readable instructions 52, the steps in the above embodiments of living body recognition based on plane detection are implemented, for example, steps S101 to S106 shown in FIG. 1. Alternatively, when the processor 50 executes the computer-readable instructions 52, the functions of the units in the foregoing device embodiments, such as the functions of the units 41 to 46 shown in FIG. 4, are implemented.
  • the computer-readable instructions 52 may be divided into one or more modules/units, and the one or more modules/units are stored in the memory 51 and executed by the processor 50, To complete this application.
  • the one or more modules/units may be a series of computer-readable instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer-readable instructions 52 in the identification terminal 5.
  • the processor 50 may be a central processing unit (Central Processing Unit, CPU), it can also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit (ASIC), Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc.
  • the general-purpose processor may be a microprocessor or the processor may also be any conventional processor or the like.
  • the memory 51 may be an internal storage unit of the identification terminal 5, such as a hard disk or memory of the identification terminal 5.
  • the memory 51 may also be an external storage device of the identification terminal 5, such as a plug-in hard disk equipped on the identification terminal 5, a smart memory card (Smart Media Card, SMC), or a Secure Digital (SD). Flash memory card Card) etc.
  • the memory 51 may also include both an internal storage unit of the identification terminal 5 and an external storage device.
  • the memory 51 is used to store the computer-readable instructions and other instructions and data required by the identification terminal 5.
  • the memory 51 can also be used to temporarily store data that has been output or will be output.
  • the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated unit can be implemented in the form of hardware or software functional unit.
  • the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.
  • the technical solution of the present application essentially or the part that contributes to the existing technology or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , Including several computer-readable instructions to enable a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application.
  • the aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disks or optical disks, and other media that can store computer-readable instructions.
  • Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
  • Volatile memory may include random access memory (RAM) or external cache memory.
  • RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Channel (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • General Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Analysis (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

一种基于平面检测的活体识别方法、装置、终端及存储介质。其中，所述活体识别方法包括：获取检测对象的平面图像(S101)；提取所述平面图像上的人脸特征点(S102)；基于提取的人脸特征点确定所述平面图像对应的整体头部姿态(S103)；将提取的人脸特征点划分为多个局部特征组，并确定每个局部特征组对应的局部头部姿态(S104)；计算所述局部头部姿态与所述整体头部姿态的姿态差异(S105)；基于所述姿态差异判断所述检测对象是否为活体(S106)。通过平面图像检测实现了活体识别，有利于提高识别效率和降低设备成本。

Description

基于平面检测的活体识别方法、装置、终端及存储介质
本申请申明享有2019年02月12日递交的申请号为201910111148.5、名称为“基于平面检测的活体识别方法、装置、终端及存储介质”中国专利申请的优先权，该中国专利申请的整体内容以参考的方式结合在本申请中。
技术领域
本申请属于活体识别技术领域，尤其涉及一种基于平面检测的活体识别方法、装置、终端及存储介质。
 背景技术
目前，活体识别主要用于在一些身份验证场景中确定检测对象的真实生理特征。例如，在人脸识别应用中，通过对眨眼、张嘴、摇头、点头等组合动作进行检测，验证当前检测对象是否为真实活体本人。可有效抵御照片、换脸、面具、遮挡以及屏幕翻拍等常见的活体攻击手段，从而进行欺诈行为的甄别，保障用户的利益。
现有的人脸活体检测方法主要存在以下问题：第一，计算时间代价高，比如，需要利用三维深度信息、采用光学流法计算人脸非刚性运行的变化，计算过程复杂；第二，需要利用额外的生物特征识别设备进行配合识别，设备成本高，比如，需要利用额外的红外人体检测设备探测检测对象的温度，或者需要利用声音采集设备联合进行声音识别等等。可见，现有技术中的活体识别方式存在计算过程复杂或者需要配备额外识别设备成本较高的问题。
 技术问题
有鉴于此，本申请提供了一种基于平面检测的活体识别方法、装置、终端及存储介质，以解决现有技术中的活体识别方式存在的计算过程复杂或者需要配备额外识别设备成本较高的问题。
 技术解决方案
本申请实施例的第一方面提供了一种基于平面检测的活体识别方法，包括：
获取检测对象的平面图像；
提取所述平面图像上的人脸特征点；
基于提取的人脸特征点确定所述平面图像对应的整体头部姿态；
将提取的人脸特征点划分为多个局部特征组，并确定每个局部特征组对应的局部头部姿态；
计算所述局部头部姿态与所述整体头部姿态的姿态差异；
基于所述姿态差异判断所述检测对象是否为活体。
 有益效果
本申请通过获取检测对象的平面图像，利用平面图像上的人脸特征点确定检测对象在该平面图像上的整体头部姿态和局部头部姿态，并根据整体头部姿态和局部头部姿态的姿态差异来判断检测对象是否为活体，本申请可以解决现有的活体识别方式存在的计算过程复杂或者需要配备额外识别设备成本较高的问题；也即，一方面，本申请的识别过程可以不涉及三维深度信息，简化了计算过程，有利于提高活体识别效率；另一方面，本申请的识别过程也无需引入额外的生物特征识别设备，降低了进行活体识别的设备成本。
 附图说明
图1为本申请实施例中基于平面检测的活体识别方法的一个实施例流程图；
图2为图1所示实施例中步骤S103的一个实施例流程图；
图3为本申请实施例中基于平面检测的活体识别方法的另一个实施例流程图；
图4为本申请实施例中基于平面检测的活体识别装置的一个实施例结构图；
图5为本申请实施例中一种识别终端的示意框图。
 本发明的实施方式
请参阅图1，为本申请实施例中一种基于平面检测的活体识别方法的一个实施例流程图，可以包括：
在步骤S101中、获取检测对象的平面图像。
在本申请实施例中，首先获取检测对象的平面图像，平面图像是指二维图像，具体可以通过配置了图像传感器的终端来获取。例如，可以通过配置有摄像头的手机终端来获取到检测对象的平面图像。
在步骤S102中、提取所述平面图像上的人脸特征点。
在本申请实施例中，人脸特征点是指在平面图像上可用于体现人脸特征的像素点或像素点集合，每个人脸特征点均可以反映人脸的一个特征；具体的，体现在平面图像上，每个人脸特征点可以为一个像素点，也可以为多个相邻像素点的集合，例如多个相邻像素点构成的一个像素块。
在本申请实施例中，提取的人脸特征点可以为预定义的多个人脸特征点，例如，提取的人脸特征点可以包括鼻尖、下巴、左眼的左角、右眼的右角、嘴的左角和嘴的右角等人脸部位在平面图像上对应的像素点或像素点集合。
在步骤S103中、基于提取的人脸特征点确定所述平面图像对应的整体头部姿态。
在本申请实施例中，对于给定的一张平面图像，其上的人物的头部姿态可以包括pitch（上下翻转，例如低头和仰头），yaw（左右翻转，例如左转或左转头部），roll（平面内旋转，例如向左或向右低头）三种角度。整体头部姿态也即人脸在三维空间的朝向，该朝向可以为上述三种三维空间内的角度之一。
在本申请实施例中，人脸在三维空间的不同朝向，反映在平面图像上，主要体现在各人脸特征点的不同的位置分布。因此，可以根据提取的各人脸特征点的位置分布来确定该平面图像上的人脸在三维空间的朝向，也即根据平面图像上人脸特征点的位置分布可以得到该平面图像对应的在三维空间的整体头部姿态。
在一种实现方式中，可以通过建立用于整体头部姿态估计的神经网络，并利用已确定头部姿态的平面图像作为训练样本对建立的神经网络进行训练和深度学习，将待确定头部姿态的平面图像输入训练好的神经网络，从而确定平面图像的整体头部姿态。
另外，在另一种实现方式中，还可以通过基于表观的方法，利用平面图像上人脸的表观特征实现整体头部姿态的估计。
可选的，如图2所示，上述步骤S103的一个实施例可以包括：
步骤S1031、获取提取的人脸特征点在所述平面图像上的位置分布，得到第一位置分布。
步骤S1032、调整预设的三维人脸模型的姿态，并在调整过程中获取所述三维人脸模型上的人脸特征点在二维平面上的投影的位置分布，得到第二位置分布。
在本申请实施例中，所述三维人脸模型上的人脸特征点与提取的所述平面图像上的人脸特征点一一对应。
步骤S1033、获取所述第二位置分布与所述第一位置分布一致时所述三维人脸模型的空间姿态，得到目标姿态。
步骤S1031、将所述目标姿态确定为所述平面图像对应的整体头部姿态。
在本申请实施例中，可以预先设置一个标准的三维人脸模型，从基准方向（例如正向）开始旋转该三维人脸模型以调整其人脸朝向，并监测该三维人脸模型上的各个人脸特征点的位置变化，当三维人脸模型上的人脸特征点在二维平面上的投影的位置分布与提取的平面图像上的人脸特征点的位置分布一致或者接近一致时，获取三维人脸模型的方位角度旋转信息，此时的方位角度旋转信息即三维人脸模型的头部姿态，也即平面图像对应的整体头部姿态。
具体的，三维人脸模型上的人脸特征点在二维平面上的投影计算，可以利用相机标定法以及直接线性变换（Direct Linear Transform，DLT）的方法实现，也即从世界坐标系（三维）到图像坐标系（二维平面）的转换。
在步骤S104中、将提取的人脸特征点划分为多个局部特征组，并确定每个局部特征组对应的局部头部姿态。
在本申请实施例中，对提取的人脸特征点进行分组，可以得到多个局部特征组，对于每个局部特征组，通过上述同样的方式调整三维人脸模型，可以得到该局部特征组对应的局部头部姿态。
可选的，上述步骤S104可以包括：
以指定数量的人脸特征点为一组将提取的人脸特征点划分为多个局部特征组；
计算每个局部特征组在所述三维人脸模型处于所述目标姿态时对应的局部空间姿态；
将每个局部特征组在所述三维人脸模型处于所述目标姿态时对应的局部空间姿态确定为该局部特征组对应的局部头部姿态。
在本申请实施例中，所述指定数量为不小于3的整数，所述指定数量小于所述提取的人脸特征点的数量，例如指定数量可以为3，也即，每三个人脸特征点作为一个局部特征组。
示例性的，将鼻尖、左眼的左角、右眼的右角三个人脸特征点划分为一组得到第一局部特征组，当三维人脸模型处于上述目标姿态时，将所述三维人脸模型上的鼻尖、左眼的左角、右眼的右角三个人脸特征点对应的平面的朝向确定为第一局部特征组对应的局部头部姿态。
同理，将下巴、左眼的左角、右眼的右角三个人脸特征点划分为一组得到第二局部特征组，当三维人脸模型处于上述目标姿态时，将所述三维人脸模型的下巴、左眼的左角、右眼的右角三个人脸特征点所在平面的朝向确定为第二局部特征组对应的局部头部姿态。
在步骤S105中、计算所述局部头部姿态与所述整体头部姿态的姿态差异。
在本申请实施例中，统计各个局部特征组对应的局部头部姿态与平面图像的整体头部姿态的姿态差异，并进一步可以通过统计的姿态差异来判断检测对象是否为活体。
进一步的，上述步骤S105可以包括：
获取上述三维人脸模型处于上述目标姿态时的姿态向量；
获取各局部特征组在所述三维人脸模型处于所述目标姿态时的平面法向量；
计算各平面法向量与所述姿态向量的夹角，其中，所述夹角的大小表示所述姿态差异的大小。
在本申请实施例中，对整体头部姿态和局部头部姿态进行向量化表示，以三维人脸模型处于上述目标姿态时的姿态向量来表示整体头部姿态，以三维人脸模型处于上述目标姿态时各局部特征组所在平面的平面法向量来表示局部头部姿态，计算各平面法向量与三维人脸模型处于上述目标姿态时的姿态向量的夹角，通过该夹角来表示局部头部姿态与整体头部姿态的姿态差异，夹角越大，表示该局部头部姿态与整体头部姿态的姿态差异越大，夹角越小，表示该局部头部姿态与整体头部姿态的姿态差异越小。
在步骤S106中、基于所述姿态差异判断所述检测对象是否为活体。
在本申请实施例中，姿态差异反映了头部的局部平面与整体姿态的差异，而实际上，对于一个活体对象，该姿态差异通常不会超过一个阈值，那么，通过将姿态差异与阈值进行比较，即可判断检测对象是否为活体，例如，当某局部特征组对应的局部头部姿态与整体头部姿态的姿态差异大于设定阈值时，可以判断所检测的对应不为活体。
进一步的，上述步骤106可以包括：
统计与所述姿态向量的夹角大于第一预设阈值的局部特征组的第一数量；
若所述第一数量大于第一指定值，则判定所述检测对象不为活体；
若所述第一数量不大于所述第一指定值，则判定所述检测对象为活体。
在本申请实施例中，为了提高识别精度，可以通过统计与所述姿态向量的夹角大于第一预设阈值的局部特征组的第一数量，当该第一数量大于第一指定值时，判定本次检测对象不为活体，当该第一数量不大于第一指定值时，判定本次检测对象为活体。
示例性的，当用平板电脑或显示器进行活体攻击时（检测对象为平板或显示器上的人物），人脸的头部姿态朝向（整体头部姿态）随着显示器平面与摄像头（采集平面图像的设备）指向之间的夹角变化会基本不变，这是因为攻击头像的视频（图像）保证的，与自然人脸相比，局部平面（局部头部姿态）与摄像头指向的夹角就会是不正确的，也即两个向量会出现夹角差异（姿态差异）。
又一示例性的，假设攻击是手持照片攻击，那么如前所述，人头部整体姿态不变，但由于纸张变形和手持倾角的变化，这些形变可以由局部平面的法向量检出。因此，本申请实施例通过计算得到的多组平面法向量（局部头部姿态）与头部总体姿态向量（整体头部姿态）的夹角，并基于该夹角的大小进行活体判定，由于真实活体目标对应的各组平面法向量与头部总体姿态向量的夹角会保持在预设阈值以内，故基于夹角的大小可以进行活体判定。
综上所述，本申请通过获取检测对象的平面图像，利用平面图像上的人脸特征点确定检测对象在该平面图像上的整体头部姿态和局部头部姿态，并根据整体头部姿态和局部头部姿态的姿态差异来判断检测对象是否为活体，本申请可以解决现有的活体识别方式存在的计算过程复杂或者需要配备额外识别设备成本较高的问题；也即，一方面，本申请的识别过程可以不涉及三维深度信息，简化了计算过程，有利于提高活体识别效率；另一方面，本申请的识别过程也无需引入额外的生物特征识别设备，降低了进行活体识别的设备成本。
请参阅图3，为本申请实施例中一种基于平面检测的活体识别方法的另一个实施例，可以包括：
步骤S301、获取检测对象的平面图像。
步骤S302、提取所述平面图像上的人脸特征点。
步骤S303、基于提取的人脸特征点确定所述平面图像对应的整体头部姿态。
步骤S304、将提取的人脸特征点划分为多个局部特征组，并确定每个局部特征组对应的局部头部姿态。
步骤S305、计算所述局部头部姿态与所述整体头部姿态的姿态差异。
在本申请实施例中，上述步骤S301至步骤S305具体可参见图1所示实施例中的步骤S101至步骤S105，在此不再赘述。
步骤S306、输出用于指示所述检测对象进行指定的头部动作的动作指示。
步骤S307、监测所述检测对象进行所述头部动作时的姿态差异变化率；
步骤S308、基于所述姿态差异变化率判断所述检测对象是否为活体。
在本申请实施例中，为了进一步提高识别的准确率，还可以在实际实施时，要求检测对象进行左右上下等一种或多种指定姿势变化，而由于攻击图像无法固定，这样可以提高算法的有效性。
在检测对象的头部进行姿势变化的动作过程中，还可以实时评估各平面法向量和头部姿态向量的夹角变化率，该夹角变化率反映了检测对象进行所述头部动作时的整体头部姿态和局部头部姿态的姿态差异变化率，根据夹角变化率也可进行活体识别判定。
在具体实施时，可以获取检测对象进行所述头部动作之后的一个平面图像，并基于该平面图像利用与上述同样的方式得到一个姿态差异，将该姿态差异与进行所述头部动作之前的平面图像对应的姿态差异进行比较，可以得到姿态差异变化率。
进一步的，上述步骤S308可以包括：
统计所述姿态差异变化率大于第二预设阈值的局部特征组的第二数量；
若所述第二数量大于第二指定值，则判定所述检测对象不为活体；
若所述第二数量不大于所述第二指定值，则判定所述检测对象为活体。
在本申请实施例中，通过统计姿态差异变化率大于第二预设阈值的局部特征组的第二数量，当该第二数量大于第二指定值时，判定本次检测对象不为活体，当该第二数量不大于第二指定值时，判定本次检测对象为活体。
综上所述，本申请通过获取检测对象的平面图像，利用平面图像上的人脸特征点确定检测对象在该平面图像上的整体头部姿态和局部头部姿态，并根据整体头部姿态和局部头部姿态的姿态差异来判断检测对象是否为活体，本申请可以解决现有的活体识别方式存在的计算过程复杂或者需要配备额外识别设备成本较高的问题；也即，一方面，本申请的识别过程可以不涉及三维深度信息，简化了计算过程，有利于提高活体识别效率；另一方面，本申请的识别过程也无需引入额外的生物特征识别设备，降低了进行活体识别的设备成本。
应理解，上述实施例中各步骤的序号的大小并不意味着执行顺序的先后，各过程的执行顺序应以其功能和内在逻辑确定，而不应对本申请实施例的实施过程构成任何限定。
对应于上文实施例所述的基于平面检测的活体识别方法，图4示出了本申请实施例提供的一种基于平面检测的活体识别装置的一个实施例结构图。
本实施例中，基于平面检测的活体识别装置4可以包括：平面图像获取单元41、特征点提取单元42，整体姿态确定单元43，局部姿态确定单元44，姿态差异计算单元45和活体判断单元46。
平面图像获取单元41，用于获取检测对象的人脸平面图像；
特征点提取单元42，用于提取所述人脸平面图像的人脸特征点；
整体姿态确定单元43，用于基于提取的人脸特征点确定所述平面图像对应的整体头部姿态；
局部姿态确定单元44，用于将提取的人脸特征点划分为多个局部特征组，并确定每个局部特征组对应的局部头部姿态；
姿态差异计算单元45，用于计算所述局部头部姿态与所述整体头部姿态的姿态差异；
活体判断单元46，用于基于所述姿态差异判断所述检测对象是否为活体。
可选的，基于平面检测的活体识别装置4还可以包括：
第一获取单元，用于获取提取的人脸特征点在所述平面图像上的位置分布，得到第一位置分布；
第二获取单元，用于调整预设的三维人脸模型的姿态，并在调整过程中获取所述三维人脸模型上的人脸特征点在二维平面上的投影的位置分布，得到第二位置分布，其中，所述三维人脸模型上的人脸特征点与提取的所述平面图像上的人脸特征点一一对应；
第三获取单元，用于获取所述第二位置分布与所述第一位置分布一致时所述三维人脸模型的空间姿态，得到目标姿态；
整体姿态确定单元43具体用于，将所述目标姿态确定为所述平面图像对应的整体头部姿态。
可选的，基于平面检测的活体识别装置4还可以包括：
特征组划分单元，用于以指定数量的人脸特征点为一组将提取的人脸特征点划分为多个局部特征组，其中，所述指定数量小于所述提取的人脸特征点的数量；
空间姿态计算单元，用于计算每个局部特征组在所述三维人脸模型处于所述目标姿态时对应的局部空间姿态；
局部姿态确定单元44具体用于，将每个局部特征组在所述三维人脸模型处于所述目标姿态时对应的局部空间姿态确定为该局部特征组对应的局部头部姿态。
可选的，基于平面检测的活体识别装置4还可以包括：
第一向量获取单元，用于获取所述三维人脸模型处于所述目标姿态时的姿态向量；
第二向量获取单元，用于获取各局部特征组在所述三维人脸模型处于所述目标姿态时的平面法向量；
姿态差异计算单元45具体用于，计算各平面法向量与所述姿态向量的夹角，其中，所述夹角的大小表示所述姿态差异的大小。
可选的，基于平面检测的活体识别装置4还可以包括：
第一数量统计单元，用于统计与所述姿态向量的夹角大于第一预设阈值的局部特征组的第一数量；
活体判断单元46具体用于，若所述第一数量大于第一指定值，则判定所述检测对象不为活体；若所述第一数量不大于所述第一指定值，则判定所述检测对象为活体。
可选的，基于平面检测的活体识别装置4还可以包括：
动作指示单元，用于输出用于指示所述检测对象进行指定的头部动作的动作指示；
差异监测单元，用于监测所述检测对象进行所述头部动作时的姿态差异变化率；
活体判断单元46还用于，基于所述姿态差异变化率判断所述检测对象是否为活体。
可选的，基于平面检测的活体识别装置4还可以包括：
第二数量统计单元，用于统计所述姿态差异变化率大于第二预设阈值的局部特征组的第二数量；
活体判断单元46具体还用于，若所述第二数量大于第二指定值，则判定所述检测对象不为活体；若所述第二数量不大于所述第二指定值，则判定所述检测对象为活体。
综上所述，本申请通过获取检测对象的平面图像，利用平面图像上的人脸特征点确定检测对象在该平面图像上的整体头部姿态和局部头部姿态，并根据整体头部姿态和局部头部姿态的姿态差异来判断检测对象是否为活体，本申请可以解决现有的活体识别方式存在的计算过程复杂或者需要配备额外识别设备成本较高的问题；也即，一方面，本申请的识别过程可以不涉及三维深度信息，简化了计算过程，有利于提高活体识别效率；另一方面，本申请的识别过程也无需引入额外的生物特征识别设备，降低了进行活体识别的设备成本。
所属领域的技术人员可以清楚地了解到，为描述的方便和简洁，上述描述的装置，模块和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在上述实施例中，对各个实施例的描述都各有侧重，某个实施例中没有详述或记载的部分，可以参见其它实施例的相关描述。
图5示出了本申请实施例提供的一种识别终端的示意框图，为了便于说明，仅示出了与本申请实施例相关的部分。
在本实施例中，所述识别终端5可以是桌上型计算机、笔记本、掌上电脑及云端服务器等计算设备。该识别终端5可包括：处理器50、存储器51以及存储在所述存储器51中并可在所述处理器50上运行的计算机可读指令52，例如执行上述的基于平面检测的活体识别方法的计算机可读指令。所述处理器50执行所述计算机可读指令52时实现上述各个基于平面检测的活体识别实施例中的步骤，例如图1所示的步骤S101至步骤S106。或者，所述处理器50执行所述计算机可读指令52时实现上述各装置实施例中各单元的功能，例如图4所示单元41至46的功能。
示例性的，所述计算机可读指令52可以被分割成一个或多个模块/单元，所述一个或者多个模块/单元被存储在所述存储器51中，并由所述处理器50执行，以完成本申请。所述一个或多个模块/单元可以是能够完成特定功能的一系列计算机可读指令段，该指令段用于描述所述计算机可读指令52在所述识别终端5中的执行过程。
所述处理器50可以是中央处理单元（Central Processing Unit，CPU），还可以是其它通用处理器、数字信号处理器（Digital Signal Processor，DSP）、专用集成电路（Application Specific Integrated Circuit，ASIC）、现场可编程门阵列（Field-Programmable Gate Array，FPGA）或者其它可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
所述存储器51可以是所述识别终端5的内部存储单元，例如识别终端5的硬盘或内存。所述存储器51也可以是所述识别终端5的外部存储设备，例如所述识别终端5上配备的插接式硬盘，智能存储卡（Smart Media Card, SMC），安全数字（Secure Digital, SD）卡，闪存卡（Flash Card）等。进一步地，所述存储器51还可以既包括所述识别终端5的内部存储单元也包括外部存储设备。所述存储器51用于存储所述计算机可读指令以及所述识别终端5所需的其它指令和数据。所述存储器51还可以用于暂时地存储已经输出或者将要输出的数据。
在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
所述集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读存储介质中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干计算机可读指令用以使得一台计算机设备（可以是个人计算机，服务器，或者网络设备等）执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器（ROM，Read-Only Memory）、随机存取存储器（RAM，Random Access Memory）、磁碟或者光盘等各种可以存储计算机可读指令的介质。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机可读指令来指令相关的硬件来完成，所述的计算机可读指令可存储于一计算机非易失性可读取存储介质中，该计算机可读指令在执行时，可包括如上述各方法的实施例的流程。其中，本申请所提供的各实施例中所使用的对存储器、存储、数据库或其它介质的任何引用，均可包括非易失性和/或易失性存储器。非易失性存储器可包括只读存储器（ROM）、可编程ROM（PROM）、电可编程ROM（EPROM）、电可擦除可编程ROM（EEPROM）或闪存。易失性存储器可包括随机存取存储器（RAM）或者外部高速缓冲存储器。作为说明而非局限，RAM以多种形式可得，诸如静态RAM（SRAM）、动态RAM（DRAM）、同步DRAM（SDRAM）、双数据率SDRAM（DDRSDRAM）、增强型SDRAM（ESDRAM）、同步链路（Synchlink） DRAM（SLDRAM）、存储器总线（Rambus）直接RAM（RDRAM）、直接存储器总线动态RAM（DRDRAM）、以及存储器总线动态RAM（RDRAM）等。
以上所述实施例仅用以说明本申请的技术方案，而非对其限制；尽管参照前述实施例对本申请进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本申请各实施例技术方案的精神和范围。

Claims (20)

  1. 一种基于平面检测的活体识别方法，其特征在于，包括：
    获取检测对象的平面图像；
    提取所述平面图像上的人脸特征点；
    基于提取的人脸特征点确定所述平面图像对应的整体头部姿态；
    将提取的人脸特征点划分为多个局部特征组，并确定每个局部特征组对应的局部头部姿态；
    计算所述局部头部姿态与所述整体头部姿态的姿态差异；
    基于所述姿态差异判断所述检测对象是否为活体。
  2. 根据权利要求1所述的基于平面检测的活体识别方法，其特征在于，所述基于提取的人脸特征点确定所述平面图像对应的整体头部姿态包括：
    获取提取的人脸特征点在所述平面图像上的位置分布，得到第一位置分布；
    调整预设的三维人脸模型的姿态，并在调整过程中获取所述三维人脸模型上的人脸特征点在二维平面上的投影的位置分布，得到第二位置分布，其中，所述三维人脸模型上的人脸特征点与提取的所述平面图像上的人脸特征点一一对应；
    获取所述第二位置分布与所述第一位置分布一致时所述三维人脸模型的空间姿态，得到目标姿态；
    将所述目标姿态确定为所述平面图像对应的整体头部姿态。
  3. 根据权利要求2所述的基于平面检测的活体识别方法，其特征在于，所述将提取的人脸特征点划分为多个局部特征组，并确定每个局部特征组对应的局部头部姿态包括：
    以指定数量的人脸特征点为一组将提取的人脸特征点划分为多个局部特征组，其中，所述指定数量小于所述提取的人脸特征点的数量；
    计算每个局部特征组在所述三维人脸模型处于所述目标姿态时对应的局部空间姿态；
    将每个局部特征组在所述三维人脸模型处于所述目标姿态时对应的局部空间姿态确定为该局部特征组对应的局部头部姿态。
  4. 根据权利要求3所述的基于平面检测的活体识别方法，其特征在于，所述计算所述局部人脸姿态与所述整体人脸姿态的姿态差异包括：
    获取所述三维人脸模型处于所述目标姿态时的姿态向量；
    获取各局部特征组在所述三维人脸模型处于所述目标姿态时的平面法向量；
    计算各平面法向量与所述姿态向量的夹角，其中，所述夹角的大小表示所述姿态差异的大小。
  5. 根据权利要求4所述的基于平面检测的活体识别方法，其特征在于，所述基于所述姿态差异判断所述检测对象是否为活体包括：
    统计与所述姿态向量的夹角大于第一预设阈值的局部特征组的第一数量；
    若所述第一数量大于第一指定值，则判定所述检测对象不为活体；
    若所述第一数量不大于所述第一指定值，则判定所述检测对象为活体。
  6. 根据权利要求1至4任一项所述的基于平面检测的活体识别方法，其特征在于，在计算所述局部人脸姿态与所述整体人脸姿态的姿态差异之后还包括：
    输出用于指示所述检测对象进行指定的头部动作的动作指示；
    监测所述检测对象进行所述头部动作时的姿态差异变化率；
    相应的，所述基于所述姿态差异判断所述检测对象是否为活体，具体为：
    基于所述姿态差异变化率判断所述检测对象是否为活体。
  7. 根据权利要求6所述的基于平面检测的活体识别方法，其特征在于，所述基于所述姿态差异变化率判断所述检测对象是否为活体，包括：
    统计所述姿态差异变化率大于第二预设阈值的局部特征组的第二数量；
    若所述第二数量大于第二指定值，则判定所述检测对象不为活体；
    若所述第二数量不大于所述第二指定值，则判定所述检测对象为活体。
  8. 一种基于平面检测的活体识别设备，其特征在于，包括：
    平面图像获取单元，用于获取检测对象的人脸平面图像；
    特征点提取单元，用于提取所述人脸平面图像的人脸特征点；
    整体姿态确定单元，用于基于提取的人脸特征点确定所述平面图像对应的整体头部姿态；
    局部姿态确定单元，用于将提取的人脸特征点划分为多个局部特征组，并确定每个局部特征组对应的局部头部姿态；
    姿态差异计算单元，用于计算所述局部头部姿态与所述整体头部姿态的姿态差异；
    活体判断单元，用于基于所述姿态差异判断所述检测对象是否为活体。
  9. 根据权利要求8所述的活体识别设备，其特征在于，基于平面检测的活体识别装置还包括：
    第一获取单元，用于获取提取的人脸特征点在所述平面图像上的位置分布，得到第一位置分布；
    第二获取单元，用于调整预设的三维人脸模型的姿态，并在调整过程中获取所述三维人脸模型上的人脸特征点在二维平面上的投影的位置分布，得到第二位置分布，其中，所述三维人脸模型上的人脸特征点与提取的所述平面图像上的人脸特征点一一对应；
    第三获取单元，用于获取所述第二位置分布与所述第一位置分布一致时所述三维人脸模型的空间姿态，得到目标姿态；
    整体姿态确定单元具体用于，将所述目标姿态确定为所述平面图像对应的整体头部姿态。
  10. 根据权利要求9所述的活体识别设备，其特征在于，所述基于平面检测的活体识别装置还包括：
    特征组划分单元，用于以指定数量的人脸特征点为一组将提取的人脸特征点划分为多个局部特征组，其中，所述指定数量小于所述提取的人脸特征点的数量；
    空间姿态计算单元，用于计算每个局部特征组在所述三维人脸模型处于所述目标姿态时对应的局部空间姿态；
    局部姿态确定单元具体用于，将每个局部特征组在所述三维人脸模型处于所述目标姿态时对应的局部空间姿态确定为该局部特征组对应的局部头部姿态。
  11. 一种终端设备，其特征在于，所述终端设备包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机可读指令，所述处理器执行时实现如下步骤：
    获取检测对象的平面图像；
    提取所述平面图像上的人脸特征点；
    基于提取的人脸特征点确定所述平面图像对应的整体头部姿态；
    将提取的人脸特征点划分为多个局部特征组，并确定每个局部特征组对应的局部头部姿态；
    计算所述局部头部姿态与所述整体头部姿态的姿态差异；
    基于所述姿态差异判断所述检测对象是否为活体。
  12. 根据权利要求11述的终端设备，其特征在于，所述基于提取的人脸特征点确定所述平面图像对应的整体头部姿态包括：
    获取提取的人脸特征点在所述平面图像上的位置分布，得到第一位置分布；
    调整预设的三维人脸模型的姿态，并在调整过程中获取所述三维人脸模型上的人脸特征点在二维平面上的投影的位置分布，得到第二位置分布，其中，所述三维人脸模型上的人脸特征点与提取的所述平面图像上的人脸特征点一一对应；
    获取所述第二位置分布与所述第一位置分布一致时所述三维人脸模型的空间姿态，得到目标姿态；
    将所述目标姿态确定为所述平面图像对应的整体头部姿态。
  13. 根据权利要求12述的生成设备，其特征在于，所述将提取的人脸特征点划分为多个局部特征组，并确定每个局部特征组对应的局部头部姿态包括：
    以指定数量的人脸特征点为一组将提取的人脸特征点划分为多个局部特征组，其中，所述指定数量小于所述提取的人脸特征点的数量；
    计算每个局部特征组在所述三维人脸模型处于所述目标姿态时对应的局部空间姿态；
    将每个局部特征组在所述三维人脸模型处于所述目标姿态时对应的局部空间姿态确定为该局部特征组对应的局部头部姿态。
  14. 根据权利要求13所述的终端设备，其特征在于，所述计算所述局部人脸姿态与所述整体人脸姿态的姿态差异包括：
    获取所述三维人脸模型处于所述目标姿态时的姿态向量；
    获取各局部特征组在所述三维人脸模型处于所述目标姿态时的平面法向量；
    计算各平面法向量与所述姿态向量的夹角，其中，所述夹角的大小表示所述姿态差异的大小。
  15. 根据权利要求14所述的终端设备，其特征在于，所述基于所述姿态差异判断所述检测对象是否为活体包括：
    统计与所述姿态向量的夹角大于第一预设阈值的局部特征组的第一数量；
    若所述第一数量大于第一指定值，则判定所述检测对象不为活体；
    若所述第一数量不大于所述第一指定值，则判定所述检测对象为活体。
  16. 一种计算机非易失性可读存储介质，所述计算机非易失性可读存储介质存储有计算机可读指令，其特征在于，所述计算机可读指令被处理器执行时实现如下步骤：
    获取检测对象的平面图像；
    提取所述平面图像上的人脸特征点；
    基于提取的人脸特征点确定所述平面图像对应的整体头部姿态；
    将提取的人脸特征点划分为多个局部特征组，并确定每个局部特征组对应的局部头部姿态；
    计算所述局部头部姿态与所述整体头部姿态的姿态差异；
    基于所述姿态差异判断所述检测对象是否为活体。
  17. 根据权利要求16所述的计算机非易失性可读存储介质，其特征在于，所述基于提取的人脸特征点确定所述平面图像对应的整体头部姿态包括：
    获取提取的人脸特征点在所述平面图像上的位置分布，得到第一位置分布；
    调整预设的三维人脸模型的姿态，并在调整过程中获取所述三维人脸模型上的人脸特征点在二维平面上的投影的位置分布，得到第二位置分布，其中，所述三维人脸模型上的人脸特征点与提取的所述平面图像上的人脸特征点一一对应；
    获取所述第二位置分布与所述第一位置分布一致时所述三维人脸模型的空间姿态，得到目标姿态；
    将所述目标姿态确定为所述平面图像对应的整体头部姿态。
  18. 根据权利要求17所述的计算机非易失性可读存储介质，其特征在于，所述将提取的人脸特征点划分为多个局部特征组，并确定每个局部特征组对应的局部头部姿态包括：
    以指定数量的人脸特征点为一组将提取的人脸特征点划分为多个局部特征组，其中，所述指定数量小于所述提取的人脸特征点的数量；
    计算每个局部特征组在所述三维人脸模型处于所述目标姿态时对应的局部空间姿态；
    将每个局部特征组在所述三维人脸模型处于所述目标姿态时对应的局部空间姿态确定为该局部特征组对应的局部头部姿态。
  19. 根据权利要求18所述的计算机非易失性可读存储介质，其特征在于，所述计算所述局部人脸姿态与所述整体人脸姿态的姿态差异包括：
    获取所述三维人脸模型处于所述目标姿态时的姿态向量；
    获取各局部特征组在所述三维人脸模型处于所述目标姿态时的平面法向量；
    计算各平面法向量与所述姿态向量的夹角，其中，所述夹角的大小表示所述姿态差异的大小。
  20. 如权利要求19所述的计算机非易失性可读存储介质，其特征在于，所述基于所述姿态差异判断所述检测对象是否为活体包括：
    统计与所述姿态向量的夹角大于第一预设阈值的局部特征组的第一数量；
    若所述第一数量大于第一指定值，则判定所述检测对象不为活体；
    若所述第一数量不大于所述第一指定值，则判定所述检测对象为活体。
PCT/CN2019/118553 2019-02-12 2019-11-14 基于平面检测的活体识别方法、装置、终端及存储介质 Ceased WO2020164284A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910111148.5 2019-02-12
CN201910111148.5A CN109977764B (zh) 2019-02-12 2019-02-12 基于平面检测的活体识别方法、装置、终端及存储介质

Publications (1)

Publication Number Publication Date
WO2020164284A1 true WO2020164284A1 (zh) 2020-08-20

Family

ID=67076912

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/118553 Ceased WO2020164284A1 (zh) 2019-02-12 2019-11-14 基于平面检测的活体识别方法、装置、终端及存储介质

Country Status (2)

Country Link
CN (1) CN109977764B (zh)
WO (1) WO2020164284A1 (zh)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112560742A (zh) * 2020-12-23 2021-03-26 杭州趣链科技有限公司 基于多尺度局部二值模式的人脸活体检测方法、装置及设备
CN113724418A (zh) * 2021-08-26 2021-11-30 广州小鹏自动驾驶科技有限公司 一种数据处理方法、装置和可读存储介质
CN114399749A (zh) * 2021-11-30 2022-04-26 际络科技(上海)有限公司 姿态基准更新方法及装置
CN114817608A (zh) * 2022-04-06 2022-07-29 福建天创信息科技有限公司 一种影像获取方法及其应用
CN115019400A (zh) * 2022-07-19 2022-09-06 北京拙河科技有限公司 违规行为检测方法及装置
CN116934848A (zh) * 2022-03-31 2023-10-24 腾讯科技（深圳）有限公司 数据处理方法、装置、设备以及介质

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109977764B (zh) * 2019-02-12 2024-12-31 平安科技（深圳）有限公司 基于平面检测的活体识别方法、装置、终端及存储介质
CN111639582B (zh) * 2020-05-26 2023-10-10 清华大学 活体检测方法及设备
CN115731589B (zh) * 2021-08-30 2026-04-07 中国移动通信有限公司研究院 一种活体识别方法、装置和电子设备
CN115239649B (zh) * 2022-07-08 2025-11-28 中国烟草总公司广西壮族自治区公司 一种图片翻拍识别方法、系统、装置及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105426827A (zh) * 2015-11-09 2016-03-23 北京市商汤科技开发有限公司 活体验证方法、装置和系统
CN105550637A (zh) * 2015-12-04 2016-05-04 小米科技有限责任公司 轮廓点定位方法及装置
WO2016082252A1 (zh) * 2014-11-27 2016-06-02 苏州福丰科技有限公司 基于云服务器的三维人脸识别机场安检方法
CN108062544A (zh) * 2018-01-19 2018-05-22 百度在线网络技术（北京）有限公司 用于人脸活体检测的方法和装置
CN109977764A (zh) * 2019-02-12 2019-07-05 平安科技（深圳）有限公司 基于平面检测的活体识别方法、装置、终端及存储介质

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107729838A (zh) * 2017-10-12 2018-02-23 中科视拓（北京）科技有限公司 一种基于深度学习的头部姿态估算方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2016082252A1 (zh) * 2014-11-27 2016-06-02 苏州福丰科技有限公司 基于云服务器的三维人脸识别机场安检方法
CN105426827A (zh) * 2015-11-09 2016-03-23 北京市商汤科技开发有限公司 活体验证方法、装置和系统
CN105550637A (zh) * 2015-12-04 2016-05-04 小米科技有限责任公司 轮廓点定位方法及装置
CN108062544A (zh) * 2018-01-19 2018-05-22 百度在线网络技术（北京）有限公司 用于人脸活体检测的方法和装置
CN109977764A (zh) * 2019-02-12 2019-07-05 平安科技（深圳）有限公司 基于平面检测的活体识别方法、装置、终端及存储介质

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112560742A (zh) * 2020-12-23 2021-03-26 杭州趣链科技有限公司 基于多尺度局部二值模式的人脸活体检测方法、装置及设备
CN113724418A (zh) * 2021-08-26 2021-11-30 广州小鹏自动驾驶科技有限公司 一种数据处理方法、装置和可读存储介质
CN113724418B (zh) * 2021-08-26 2023-07-04 广州小鹏自动驾驶科技有限公司 一种数据处理方法、装置和可读存储介质
CN114399749A (zh) * 2021-11-30 2022-04-26 际络科技(上海)有限公司 姿态基准更新方法及装置
CN116934848A (zh) * 2022-03-31 2023-10-24 腾讯科技（深圳）有限公司 数据处理方法、装置、设备以及介质
CN114817608A (zh) * 2022-04-06 2022-07-29 福建天创信息科技有限公司 一种影像获取方法及其应用
CN115019400A (zh) * 2022-07-19 2022-09-06 北京拙河科技有限公司 违规行为检测方法及装置
CN115019400B (zh) * 2022-07-19 2023-03-03 北京拙河科技有限公司 违规行为检测方法及装置

Also Published As

Publication number Publication date
CN109977764A (zh) 2019-07-05
CN109977764B (zh) 2024-12-31

Similar Documents

Publication Publication Date Title
WO2020164284A1 (zh) 基于平面检测的活体识别方法、装置、终端及存储介质
US11551481B2 (en) Living body detection method and apparatus, electronic device, storage medium, and related system to which living body detection method is applied
EP3614303B1 (en) Method and apparatus for facial recognition
US10339402B2 (en) Method and apparatus for liveness detection
US10268875B2 (en) Method and apparatus for registering face, and method and apparatus for recognizing face
JP6528764B2 (ja) 顔照合装置、方法、及び、記録媒体
CN106203248A (zh) 用于脸部识别的方法和设备
JP7064257B2 (ja) 画像深度確定方法及び生き物認識方法、回路、装置、記憶媒体
JP6071002B2 (ja) 信頼度取得装置、信頼度取得方法および信頼度取得プログラム
CN109766785A (zh) 一种人脸的活体检测方法及装置
CN112418153B (zh) 图像处理方法、装置、电子设备和计算机存储介质
WO2023034251A1 (en) Spoof detection based on challenge response analysis
KR20160093523A (ko) 안구 위치 정보 확정 방법 및 장치
CN113298158B (zh) 数据检测方法、装置、设备及存储介质
WO2021197466A1 (zh) 眼球检测方法、装置、设备及存储介质
WO2019033575A1 (zh) 电子装置、人脸追踪的方法、系统及存储介质
CN108288023B (zh) 人脸识别的方法和装置
CN110188630A (zh) 一种人脸识别方法和相机
CN112017212A (zh) 人脸关键点跟踪模型的训练、跟踪方法及系统
CN111563490B (zh) 一种人脸关键点跟踪方法、装置及电子设备
WO2018103416A1 (zh) 用于人脸图像的检测方法和装置
CN113128427A (zh) 人脸识别方法、装置、计算机可读存储介质及终端设备
US20250029425A1 (en) Live human face detection method and apparatus, computer device, and storage medium
CN104636727A (zh) 一种适应多表情多姿态的人脸识别方法
CN110046554B (zh) 一种人脸对齐方法和相机

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19914979

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS (EPO FORM 1205A DATED 05.10.2021)

122 Ep: pct application non-entry in european phase

Ref document number: 19914979

Country of ref document: EP

Kind code of ref document: A1