WO2020164266A1 - 一种活体检测方法、系统及终端设备 - Google Patents

一种活体检测方法、系统及终端设备 Download PDF

Info

Publication number
WO2020164266A1
WO2020164266A1 PCT/CN2019/117188 CN2019117188W WO2020164266A1 WO 2020164266 A1 WO2020164266 A1 WO 2020164266A1 CN 2019117188 W CN2019117188 W CN 2019117188W WO 2020164266 A1 WO2020164266 A1 WO 2020164266A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
key area
depth
key
focus
Prior art date
Application number
PCT/CN2019/117188
Other languages
English (en)
French (fr)
Inventor
蒋诗彬
陆进
陈斌
宋晨
Original Assignee
平安科技（深圳）有限公司
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 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2020164266A1 publication Critical patent/WO2020164266A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition

Definitions

  • This application belongs to the field of computer technology, and in particular relates to a living body detection method, system and terminal device.
  • face recognition systems are increasingly used in scenarios that require identity verification in the fields of security and finance, such as remote bank account opening, access control systems, and remote transaction operation verification.
  • identity verification in the fields of security and finance, such as remote bank account opening, access control systems, and remote transaction operation verification.
  • the above-mentioned technology to determine that the person being verified is a legal living organism is called a living body detection technology, and its purpose is to determine whether the acquired biological characteristics come from a living, on-site, and real person.
  • Existing living body detection technologies usually need to rely on special hardware equipment, such as infrared cameras, depth cameras and other image acquisition equipment or complex active light sources such as DLP projectors to prevent simple photos, 3D Attacks such as face models or masks.
  • the current face recognition system has the problem that it cannot perform live body detection based on traditional cameras, and requires additional hardware equipment.
  • the embodiments of the present application provide a living body detection method, system, and terminal device to solve the problem that the current face recognition system cannot perform living body detection based on traditional cameras and requires additional hardware equipment.
  • the first aspect of this application provides a living body detection method, including:
  • the detection object is a living body.
  • the second aspect of the present application provides a living body detection system, including:
  • the collection module is used to collect the image of the detected object and locate the key area according to the facial features and the classifier;
  • An estimation module configured to set all the key regions of the image of the detection object as a key region group, and estimate the depth of each pixel of each key region in the key region group;
  • the shooting module is used to separately focus and shoot each key area based on the estimated depth, and obtain a focus image of each key area;
  • the judgment module is used for judging whether the detection object is a living body according to the image quality of the in-focus image of each key area.
  • the third aspect of the present application provides a terminal device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor.
  • a terminal device including a memory, a processor, and a computer program stored in the memory and capable of running on the processor.
  • the detection object is a living body.
  • the fourth aspect of the present application provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps are implemented:
  • the detection object is a living body.
  • the face recognition method, system, and terminal device provided by the present application implement live body detection by using the focus capability of a traditional camera to perform depth-of-field measurement, which can achieve effective live body detection without adding depth camera equipment, effectively reducing costs, It solves the problem that the current face recognition system cannot perform live body detection based on traditional cameras and needs to add hardware devices.
  • FIG. 1 is a schematic diagram of the implementation process of a living body detection method provided in Embodiment 1 of the present application;
  • step S102 is a schematic diagram of the implementation process of step S102 corresponding to Embodiment 1 provided in Embodiment 2 of the present application;
  • FIG. 3 is a schematic diagram of the implementation process of step S103 corresponding to Embodiment 1 provided in Embodiment 3 of the present application;
  • step S104 is a schematic diagram of the implementation process of step S104 corresponding to Embodiment 1 provided by Embodiment 4 of the present application;
  • FIG. 5 is a schematic structural diagram of a living body detection system provided by Embodiment 5 of the present application.
  • FIG. 6 is a schematic structural diagram of the timing estimation module 102 in the fifth embodiment provided by the sixth embodiment of the present application.
  • FIG. 7 is a schematic structural diagram of the photographing module 103 in the fifth embodiment provided by the seventh embodiment of the present application.
  • FIG. 8 is a schematic structural diagram of the judgment module 104 in the fifth embodiment provided by the eighth embodiment of the present application
  • FIG. 9 is a schematic diagram of a terminal device provided in Embodiment 9 of the present application.
  • the embodiments of this application provide a live body detection method, system and terminal device, which use the focus ability of traditional cameras to perform depth of field
  • the measurement method is used to realize live body detection, which can realize effective live body detection without adding depth camera equipment, effectively reduce the cost, and solve the problem that the current face recognition system cannot perform live body detection based on traditional cameras and requires additional hardware equipment.
  • this embodiment provides a living body detection method, which specifically includes:
  • Step S101 Collect an image of the detection object and locate the key area according to the facial features of the face and the classifier.
  • the camera module of the face recognition system captures the detected object, obtains the captured image, performs grayscale transformation and filtering on the image captured by the camera module, and obtains the grayscale image of the image;
  • the integral quickly calculates the Harr-Like wavelet feature value, and applies it to the offline trained AdaBoost-Cascade classifier to determine whether the grayscale image contains the face area; according to the facial features of the face and the AdaBoost-Cascade classifier
  • the area performs key area positioning and determines the specific position coordinates of each key area in the above image.
  • the above key areas include facial feature areas such as eyes, eyebrows, nose, mouth, and jaw.
  • the above-mentioned key area may also include a background area near the human face area.
  • Step S102 Set all the key regions of the image of the detection object as a key region group, and estimate the depth of each pixel point of each key region in the key region group.
  • the key area group is set according to each key area of the image of the detected object actually taken.
  • the living body detection method provided in this embodiment mainly uses the focusing ability of a traditional camera to estimate the depth. Therefore, when setting the key area group, it is necessary to comprehensively consider the key areas of the head of the 3D stereo Position, set all key areas such as the eyes, eyebrows, nose, mouth, and chin, and the background area near the face area based on the facial features of the face and the classifier as the key area group.
  • the neural network model is used to estimate the depth of each key area of the key area group, and the depth of each pixel of the image is estimated through the neural network model, and then the specific location of each key area is obtained according to the specific location of each key area. The depth of the pixel.
  • Step S103 focusing and shooting each key area respectively based on the estimated depth, and acquiring a focus image of each key area.
  • focus shooting is performed sequentially according to the depth of each key area, and then the focus image of each key area is obtained.
  • the focus stack depth estimation method is used to determine the focus point on the depth surface of each key area, and the shooting module is controlled to focus and shoot the focus point of each key area of the key area group, and it can be obtained The focus image corresponding to each key area of the key area group.
  • Step S104 Determine whether the detection object is a living body according to the image quality of the in-focus image of each key area.
  • the degree of blur of each focus image is acquired, the image quality of the focus image is determined based on the degree of blur, and the determination threshold is set based on the degree of blur. If the degree of blur of each focus image is higher than the determination threshold, the detection object It is non-living.
  • the image contrast of each focused image is acquired, the image quality of the focused image is determined based on the image contrast, and the determination threshold is set based on the image contrast. If the image contrast of each focused image is higher than the determination threshold, the detection object It is non-living.
  • the exposure level of each focus image is acquired, the image quality of the focus image is determined based on the exposure level, the determination threshold is set based on the exposure level of the image, and the exposure level of each focus image is lower than the determination threshold, then the The detection object is a non-living body.
  • the aforementioned face recognition method further includes the following steps after step S102:
  • Step S105 It is judged whether the depth of the center point of each key region obtained by estimation is the same.
  • Step S106 If the depths of the center points of the key regions are the same, it is determined that the detection object is a non-living body.
  • plane attack objects such as screen recording and paper shooting
  • the focus images obtained after focusing and shooting in different key areas estimate the depth information of the focus images corresponding to each key area, and determine whether the depth of the focus image corresponding to each key area is the same. If the depth of the focus image corresponding to each key area is the same The same means that the detection object is a flat object, and therefore the detection object is a non-living body. If the depths of the focus images corresponding to the key regions are not the same, steps S103 to S104 are executed to further determine whether the detection object is a living body.
  • each key area refers to the most central position of each key area.
  • the center point of the key area is the intersection of two diagonals of the rectangle. If the key area is set as a circular area, the center point of the key area is the center of the circle.
  • the living body detection method provided in this embodiment realizes living body detection by using the focus capability of a traditional camera to perform depth-of-field measurement. It can realize effective living body detection without adding depth camera equipment, effectively reducing costs, and solving the current face recognition
  • the system has the problem that it cannot perform live detection based on traditional cameras and needs to add hardware devices.
  • step S102 in the first embodiment specifically includes:
  • Step S201 Constructing and training a neural network model of estimated depth.
  • the above-mentioned neural network model for predicting the depth can be a Multi-Scale deep neural network model, a VGG16 neural network model, or a deep neural network model based on Fully Convolutional Networks, which is not limited here. . Since the above-mentioned Multi-Scale deep neural network model, VGG16 neural network model, and deep neural network model based on Fully Convolutional Networks are existing deep neural network modules, the specific construction and training process will not be repeated.
  • Step S202 Input the collected image into the neural network model of the estimated depth to obtain the estimated depth of the image.
  • the depth of an image includes the depth of each pixel of the image.
  • Step S203 Determine the depth of each pixel of each key area of the key area group according to the position of each key area and the depth of the image.
  • the depth of each pixel in each key area is determined according to the specific position of each key area located.
  • the key area is positioned by a rectangular frame, and the position information of the key area is determined by locating the rectangular frame corresponding to the key area by the coordinates of the four corners.
  • the depth of each pixel in each key area is extracted according to the determined position information.
  • step S103 in Embodiment 1 specifically includes:
  • Step S301 Determine the focal point on the depth plane of each key area by using the focus stack depth estimation method according to the depth of each key area.
  • using the focus stack depth estimation method to determine the focus point of the camera on the depth surface of each key area is specifically: calculating the gradient of each pixel point of the focus stack according to the depth of each pixel point in the key area, and the gradient image Perform average filtering, then calculate the square sum of the gradient values of the three channels (R channel, G channel and B channel) and take the average, and then root the square sum average to get the gradient value of the key area and the pixel at the same position in the focus stack
  • the maximum gradient indicates that the pixel position is the focal point of the depth surface.
  • Step S302 Control the camera to focus and photograph the focus points of each key area of the key area group to obtain the focus image of each key area.
  • the image obtained by shooting at the focal point is the clearest image captured by the depth surface, so the image captured by controlling the shooting module to focus on the pixel (focus point) is the focus of the key area image.
  • the image captured by controlling the shooting module to focus on the pixel (focus point) is the focus of the key area image.
  • step S104 in Embodiment 1 specifically includes:
  • Step S401 Obtain the blur degree of the in-focus image of each key area.
  • the Laplacian variance algorithm is used to determine the blur degree of the focus image in each key area.
  • the grayscale transformation is performed on the focus image of each key area to obtain the grayscale image of the focus image of each key area
  • the Laplacian mask is used for the convolution operation on the grayscale image, and then the variance is calculated to obtain the The degree of blur in the focused image.
  • the degree of blur is used to measure the quality of the captured image. The higher the degree of blur, the worse the quality of the corresponding image.
  • the above Laplacian mask is:
  • the degree of blurring of each focused image can also be obtained by using a neural network model.
  • Inputting the focused image into the neural network model for obtaining the degree of blurring can output the degree of blurring of the focused image.
  • the neural network model for obtaining the degree of blur can be the VGG16 neural network model, and its construction and training process will not be repeated here.
  • Step S402 Calculate the blur degree of the image according to the blur degree of the focused image in each key area.
  • the fuzzy degree value of the image is comprehensively calculated based on the proportion of the fuzzy degree of the key area. The calculation is:
  • Blur is the blur degree value of the image
  • Pi is the specific gravity coefficient of the i-th key region
  • the specific gravity coefficient of the key area is set according to the object corresponding to the key area, and the specific gravity coefficient can be set based on experience. For example, the specific gravity coefficient of the key area containing the eyes is greater than the key area containing the eyebrows The specific gravity coefficient. In practical applications, the specific gravity coefficients of each key area are given in advance based on the delineated key areas. It should be noted that the sum of the specific gravity coefficients of the key areas is 1.
  • Step S403 Determine whether the blur degree of the image exceeds a determination threshold.
  • Step S404 If the blur degree of the image exceeds the judgment threshold, determine that the detection object is a non-living body.
  • the image quality of multiple in-focus images is evaluated by judging whether the blur degree of the image exceeds a preset judgment threshold to determine whether the detection object is a living body, and when the blur degree of the image exceeds the judgment threshold, It is determined that the detection object is a non-living body, and if the blur degree of the image does not exceed the determination threshold, the detection object is determined to be a living body.
  • this embodiment provides a living body detection system 100 for performing the method steps in the first embodiment, which includes an acquisition module 101, an estimation module 102, a photographing module 103, and a judgment module 104.
  • the acquisition module 101 is used to acquire an image of a detection object and locate key areas according to the facial features and the classifier.
  • the estimation module 102 is configured to set all key regions of the image of the detection object as a key region group, and estimate the depth of each pixel point of each key region in the key region group.
  • the photographing module 103 is configured to separately focus and photograph each key area based on the estimated depth, and obtain a focus image of each key area.
  • the judging module 104 is configured to judge whether the detection object is a living body according to the image quality of the in-focus image of each key area.
  • the aforementioned face recognition system 100 further includes a depth judgment module.
  • the above-mentioned depth judgment module is used to judge whether the depths of the center points of the key regions obtained by estimation are the same. If the depths of the center points of the key regions are the same, it is determined that the detection object is a non-living body.
  • the living body detection system provided by the embodiment of the present application is based on the same concept as the method embodiment shown in FIG. 1 of the present application, and the technical effect brought by it is the same as the method embodiment shown in FIG. 1 of the present application.
  • the living body detection system provided by this embodiment can also realize living body detection by using the focus ability of a traditional camera to perform depth-of-field measurement. It can realize effective living body detection without adding depth camera equipment, effectively reducing costs, and solving In addition, the current face recognition system has the problem that it cannot perform live detection based on traditional cameras and requires additional hardware equipment.
  • Embodiment 6 is a diagrammatic representation of Embodiment 6
  • the estimation module 102 in the fifth embodiment includes a structure for executing the method steps in the embodiment corresponding to FIG. 2, which includes a construction unit 201, an input unit 202, and a determination unit 203.
  • the construction unit 201 is used to construct and train a neural network model of the estimated depth.
  • the input unit 202 is used to input the collected image into the neural network model of the estimated depth to obtain the estimated depth of the image, and the depth of the image includes the depth of each pixel of the image.
  • the determining unit 203 is configured to determine the depth of each pixel of each key area of the key area group according to the position of each key area and the depth of the image.
  • the photographing module 103 in the fifth embodiment includes a structure for executing the method steps in the embodiment corresponding to FIG. 3, which includes a focus point determining unit 301 and a photographing unit 302.
  • the focus point determination unit 301 is used to determine the focus point on the depth plane of each key area by using the focus stack depth estimation method according to the depth of each key area.
  • the photographing unit 302 is used to control the camera to respectively focus and photograph the focus points of each key area of the key area group to obtain a focused image of each key area.
  • Embodiment 8 is a diagrammatic representation of Embodiment 8
  • the judgment module 104 in the fifth embodiment includes a structure for executing the method steps in the embodiment corresponding to FIG. 4, which includes an acquisition unit 401, a calculation unit 402, and a judgment unit 403.
  • the obtaining unit 401 is used to obtain the blur degree of the in-focus image of each key area.
  • the calculation unit 402 is configured to calculate the blur degree of the image according to the blur degree of the focus image of each key area.
  • the judging unit 403 is used to judge whether the blur degree of the image exceeds the judgment threshold; if the blur degree of the image exceeds the judgment threshold, it is determined that the detection object is an inanimate body.
  • FIG. 9 is a schematic diagram of a terminal device provided in Embodiment 7 of the present application.
  • the terminal device 9 of this embodiment includes a processor 90, a memory 91, and a computer program 92, such as a program, stored in the memory 91 and running on the processor 90.
  • the processor 90 executes the computer program 92, the steps in the foregoing method embodiments, such as steps S101 to S104 shown in FIG. 1, are implemented.
  • the processor 90 executes the computer program 92, the functions of the modules/units in the foregoing system embodiment, such as the functions of the modules 101 to 104 shown in FIG. 5, are realized.
  • the computer program 92 may be divided into one or more modules/units, and the one or more modules/units are stored in the memory 91 and executed by the processor 90 to complete This application.
  • the one or more modules/units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program 92 in the terminal device 9.
  • the computer program 92 can be divided into an acquisition module, an estimation module, a photographing module, and a judgment module. The specific functions of each module are as follows:
  • the collection module is used to collect the image of the detected object and locate the key area according to the facial features and the classifier;
  • An estimation module configured to set all the key regions of the image of the detection object as a key region group, and estimate the depth of each pixel of each key region in the key region group;
  • the shooting module is used to separately focus and shoot each key area based on the estimated depth, and obtain a focus image of each key area;
  • the judgment module is used for judging whether the detection object is a living body according to the image quality of the in-focus image of each key area.
  • the terminal device 9 may be a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud management server.
  • the terminal device may include, but is not limited to, a processor 90 and a memory 91.
  • FIG. 9 is only an example of the terminal device 9 and does not constitute a limitation on the terminal device 9. It may include more or less components than shown in the figure, or a combination of certain components, or different components.
  • the terminal device may also include input and output devices, network access devices, buses, etc.
  • the so-called processor 90 may be a central processing unit (Central Processing Unit, CPU), other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (ASIC), Ready-made programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate 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 91 may be an internal storage unit of the terminal device 9, such as a hard disk or memory of the terminal device 9.
  • the memory 91 may also be an external storage device of the terminal device 9, for example, a plug-in hard disk equipped on the terminal device 9, a smart memory card (Smart Media Card, SMC), and a Secure Digital (SD) Card, Flash Card, etc. Further, the memory 91 may also include both an internal storage unit of the terminal device 9 and an external storage device.
  • the memory 91 is used to store the computer program and other programs and data required by the terminal device.
  • the memory 91 can also be used to temporarily store data that has been output or will be output.
  • system/terminal device and method may be implemented in other ways.
  • the system/terminal device embodiments described above are only illustrative.
  • the division of the modules or units is only a logical function division, and there may be other divisions in actual implementation, such as multiple units.
  • components can be combined or integrated into another system, or some features can be omitted or not implemented.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, systems or units, and may be in electrical, mechanical or other forms.
  • the unit described as a separate component may or may not be physically separated, and the component displayed as a unit may or may not be a physical unit, that is, it may be located in one place, or may be distributed to multiple network units on. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
  • the functional units in the 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 module/unit is implemented in the form of a software functional unit and set as an independent product for sale or use, it can be stored in a computer readable storage medium.
  • this application implements all or part of the processes in the above-mentioned embodiments and methods, and can also be completed by instructing relevant hardware through a computer program.
  • the computer program can be stored in a computer-readable storage medium. When the program is executed by the processor, the steps of the foregoing method embodiments can be implemented.
  • the computer program includes computer program code, and the computer program code may be in the form of source code, object code, executable file, or some intermediate forms.
  • the computer-readable medium may include: any entity or system capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory) , Random Access Memory (RAM, Random Access Memory), electrical carrier signal, telecommunications signal, and software distribution media, etc.
  • ROM Read-Only Memory
  • RAM Random Access Memory
  • electrical carrier signal telecommunications signal
  • software distribution media etc.
  • the content contained in the computer-readable medium can be appropriately added or deleted in accordance with the requirements of the legislation and patent practice in the jurisdiction.
  • the computer-readable medium Does not include electrical carrier signals and telecommunication signals.

Abstract

一种人脸识别方法、系统及终端设备，所述方法包括：采集检测对象的图像并根据人脸的五官形状特征及分类器定位关键区域（S101）；将所述检测对象的图像的所有关键区域设置为关键区域组，估算所述关键区域组中各关键区域的各个像素点的深度（S102）；基于估算得到的深度分别对各关键区域进行对焦拍摄，获取各关键区域的对焦图像（S103）；根据所述各关键区域的对焦图像的图像质量判断检测对象是否为活体（S104），所述方法通过利用传统相机的对焦能力进行景深测量的方法来实现活体检测，无需增加深度摄像设备就能实现有效地活体检测，有效降低成本，解决了目前的人脸识别系统存在无法基于传统相机进行活体检测，需要增加硬件设备的问题。

Description

一种活体检测方法、系统及终端设备
本申请申明享有2019年2月13日递交的申请号为201910112917.3、名称为“一种活体检测方法、系统及终端设备”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
本申请属于计算机技术领域，尤其涉及一种活体检测方法、系统及终端设备。
背景技术
目前，人脸识别系统越来越多地应用于安防、金融等领域中需要身份验证的场景，诸如银行远程开户、门禁系统、远程交易操作验证等。在这些高安全级别的应用领域中，除了确保被验证者的人脸相似度符合数据库中存储的底库数据外，首先需要确定被验证者是一个合法的生物活体。上述确定被验证者为是一个合法的生物活体的技术被称为活体检测技术，其目的是判断获取到的生物特征是否来自一个有生命、在现场的、真实的人。目前还没有公认成熟的活体验证方案，已有的活体检测技术通常需要依赖特殊的硬件设备，诸如红外相机、深度相机等图像采集设备或者诸如DLP投影仪的复杂的主动光源来防范简单照片、3D人脸模型或者面具等方式的攻击。
综上所述，目前的人脸识别系统存在无法基于传统相机进行活体检测，需要增加硬件设备的问题。
技术问题
有鉴于此，本申请实施例提供了一种活体检测方法、系统及终端设备，以解决目前的人脸识别系统存在无法基于传统相机进行活体检测，需要增加硬件设备的问题。
技术解决方案
本申请的第一方面提供了一种活体检测方法，包括：
采集检测对象的图像并根据人脸的五官形状特征及分类器定位关键区域；
将所述检测对象的图像的所有关键区域设置为关键区域组，估算所述关键区域组中各关键区域的各个像素点的深度；
基于估算得到的深度分别对各关键区域进行对焦拍摄，获取各关键区域的对焦图像；
根据所述各关键区域的对焦图像的图像质量判断检测对象是否为活体。
本申请的第二方面提供了一种活体检测系统，包括：
采集模块，用于采集检测对象的图像并根据人脸的五官形状特征及分类器定位关键区域；
估算模块，用于将所述检测对象的图像的所有关键区域设置为关键区域组，估算所述关键区域组中各关键区域的各个像素点的深度；
拍摄模块，用于基于估算得到的深度分别对各关键区域进行对焦拍摄，获取各关键区域的对焦图像；
判断模块，用于根据所述各关键区域的对焦图像的图像质量判断检测对象是否为活体。
本申请的第三方面提供了一种终端设备，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程序，所述处理器执行所述计算机程序时实现以下步骤：
采集检测对象的图像并根据人脸的五官形状特征及分类器定位关键区域；
将所述检测对象的图像的所有关键区域设置为关键区域组，估算所述关键区域组中各关键区域的各个像素点的深度；
基于估算得到的深度分别对各关键区域进行对焦拍摄，获取各关键区域的对焦图像；
根据所述各关键区域的对焦图像的图像质量判断检测对象是否为活体。
本申请的第四方面提供了一种计算机可读存储介质，所述计算机可读存储介质存储有计算机程序，所述计算机程序被处理器执行时实现以下步骤：
采集检测对象的图像并根据人脸的五官形状特征及分类器定位关键区域；
将所述检测对象的图像的所有关键区域设置为关键区域组，估算所述关键区域组中各关键区域的各个像素点的深度；
基于估算得到的深度分别对各关键区域进行对焦拍摄，获取各关键区域的对焦图像；
根据所述各关键区域的对焦图像的图像质量判断检测对象是否为活体。
有益效果
本申请提供的一种人脸识别方法、系统及终端设备，通过利用传统相机的对焦能力进行景深测量的方法来实现活体检测，无需增加深度摄像设备就能实现有效地活体检测，有效降低成本，解决了目前的人脸识别系统存在无法基于传统相机进行活体检测，需要增加硬件设备的问题。
附图说明
为了更清楚地说明本申请实施例中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1是本申请实施例一提供的一种活体检测方法的实现流程示意图；
图2是本申请实施例二提供的对应实施例一步骤S102的实现流程示意图；
图3是本申请实施例三提供的对应实施例一步骤S103的实现流程示意图；
图4是本申请实施例四提供的对应实施例一步骤S104的实现流程示意图；
图5是本申请实施例五提供的一种活体检测系统的结构示意图；
图6是本申请实施例六提供的对应实施例五中定时估算模块102的结构示意图；
图7是本申请实施例七提供的对应实施例五中拍摄模块103的结构示意图；
图8是本申请实施例八提供的对应实施例五中判断模块104的结构示意图
图9是本申请实施例九提供的终端设备的示意图。
本发明的实施方式
以下描述中，为了说明而不是为了限定，提出了诸如特定系统结构、技术之类的具体细节，以便透彻理解本申请实施例。然而，本领域的技术人员应当清楚，在没有这些具体细节的其它实施例中也可以实现本申请。在其它情况中，省略对众所周知的系统、系统、电路以及方法的详细说明，以免不必要的细节妨碍本申请的描述。
本申请实施例为了解决目前的人脸识别系统存在无法基于传统相机进行活体检测，需要增加硬件设备的问题，提供了一种活体检测方法、系统及终端设备，通过利用传统相机的对焦能力进行景深测量的方法来实现活体检测，无需增加深度摄像设备就能实现有效地活体检测，有效降低成本，解决了目前的人脸识别系统存在无法基于传统相机进行活体检测，需要增加硬件设备的问题。
为了说明本申请所述的技术方案，下面通过具体实施例来进行说明。
实施例一：
如图1所示，本实施例提供了一种活体检测方法，其具体包括：
步骤S101：采集检测对象的图像并根据人脸的五官形状特征及分类器定位关键区域。
在具体应用中，通过人脸识别系统的摄像模块拍摄检测对象，获取拍摄到的图像，对摄像模块采集到图像进行灰度变换和滤波处理，获得该图像的灰度图；对灰度图利用积分快速计算出Harr-Like小波特征值，应用到离线训练好的AdaBoost-Cascade分类器，判别该灰度图像是否包含人脸区域；根据人脸的五官形状特征和AdaBoost-Cascade分类器对人脸区域进行关键区域定位，确定各关键区域在上述图像的具体位置坐标。需要说明的是，上述关键区域包括双眼、双眉、鼻子、嘴巴和下颚等人脸特征区域。还需要说明的是，为了区分人脸和背景，上述关键区域还可以包括人脸区域附近的背景区域。
步骤S102：将所述检测对象的图像的所有关键区域设置为关键区域组，估算所述关键区域组中各关键区域的各个像素点的深度。
在具体应用中，首先根据实际拍摄到的检测对象的图像的各个关键区域来设置关键区域组。需要说明的是，本实施例所提供的活体检测方法主要是通过利用传统相机的对焦能力来进行深度估算，因此在设置关键区域组时需要综合考虑3D立体的人的头部的各个关键区域的位置，将根据人脸的五官形状特征及分类器定位到的双眼、双眉、鼻子、嘴巴和下颚等人脸特征区域以及人脸区域附近的背景区域等所有关键区域设置为关键区域组。
在具体应用中，采用神经网络模型估算关键区域组的各关键区域的深度，通过神经网络模型估算图像的各个像素点的深度，再根据定位到的各个关键区域的具体位置获取各个关键区域的各个像素点的深度。
步骤S103：基于估算得到的深度分别对各关键区域进行对焦拍摄，获取各关键区域的对焦图像。
在具体应用中，依次根据各关键区域的深度进行对焦拍摄，进而获取到各关键区域的对焦图像。对关键区域组的各个关键区域都采用焦点堆栈深度估计法确定各个关键区域的深度面上的聚焦点，控制拍摄模块分别对关键区域组的各个关键区域的聚焦点进行对焦拍摄，就能够获取到该关键区域组的各关键区域对应的对焦图像。
步骤S104：根据所述各关键区域的对焦图像的图像质量判断检测对象是否为活体。
在具体应用中，获取各个对焦图像的模糊程度，基于模糊程度判断所述对焦图像的图像质量，基于模糊程度设置判定阈值，各个对焦图像的模糊程度高于所述判定阈值，则所述检测对象为非活体。
在具体应用中，获取各个对焦图像的图像对比度，基于图像对比度判断所述对焦图像的图像质量，基于图像对比度设置判定阈值，各个对焦图像的图像对比度高于所述判定阈值，则所述检测对象为非活体。
在具体应用中，获取各个对焦图像的曝光程度，基于曝光程度判断所述对焦图像的图像质量，基于图像的曝光程度设置判定阈值，各个对焦图像的曝光程度低于所述判定阈值，则所述检测对象为非活体。
在一个实施例中，上述人脸识别方法，在步骤S102之后还包括以下步骤：
步骤S105：判断估算得到的各关键区域的中心点的深度是否相同。
步骤S106：若个各关键区域的的中心点的深度相同，则确定所述检测对象为非活体。
在具体应用中，对于平面攻击对象，如录屏和纸张拍摄等攻击对象的检测，通过对图像进行平面检测就能够判断其是否为平面对象。通过判断不同的关键区域进行对焦拍摄后获取的对焦图像，估计各关键区域对应的对焦图像的深度信息，判断各关键区域对应的对焦图像的深度是否相同，若各关键区域对应的对焦图像的深度相同，则表明检测对象为平面对象，因此该检测对象为非活体。若各关键区域对应的对焦图像的深度不相同，则执行步骤S103至S104，以进一步判断检测对象是否为活体。需要说明的是，各关键区域的中心点是指各个关键区域最中心的位置，示例性的，若设置关键区域为矩形区域，则关键区域的中心点为矩形的两条对角线的交点，若设置关键区域为圆形区域，则关键区域的中心点为圆形的圆心。
本实施例提供的活体检测方法，通过利用传统相机的对焦能力进行景深测量的方法来实现活体检测，无需增加深度摄像设备就能实现有效地活体检测，有效降低成本，解决了目前的人脸识别系统存在无法基于传统相机进行活体检测，需要增加硬件设备的问题。
实施例二：
如图2所示，在本实施例中，实施例一中的步骤S102具体包括：
步骤S201：构建和训练预估深度的神经网络模型。
在具体应用中，上述预估深度的神经网络模型可以为Multi-Scale的深度神经网络模型，也可以为VGG16神经网络模型，还可以为基于Fully Convolutional Networks的深度神经网 络模型，在此不加以限制。由于上述Multi-Scale的深度神经网络模型、VGG16神经网络模型、基于Fully Convolutional Networks的深度神经网络模型作为现有的深度神经网络模块，因此不对其具体的构建和训练过程再加以赘述。
步骤S202：将采集到的图像输入到预估深度的神经网络模型中，以获取估算的所述图像的深度。
在具体应用中，图像的深度包括图像各个像素点的深度。
在具体应用中，当构建和训练完用于预估图像的深度信息的神经网络模型后，将采集到的图像输入到该神经网络模型中就能输出该图像的深度信息。
步骤S203：根据各关键区域的位置及所述图像的深度确定关键区域组的各关键区域的各个像素点的深度。
在具体应用中，根据定位到的各个关键区域的具体位置来确定各个关键区域的各个像素点的深度。示例性的，假设关键区域通过矩形边框来实现定位，通过定位到该关键区域对应的矩形边框的是四角的坐标，确定该关键区域的位置信息，在获得该图像的各个像素点的深度后，根据确定的位置信息提取各个关键区域的各个像素点的深度。
实施例三：
如图3所示，在本实施例中，实施例一中的步骤S103具体包括：
步骤S301：根据各关键区域的深度利用焦点堆栈深度估计法确定在各关键区域的深度面上的聚焦点。
在具体应用中，利用焦点堆栈深度估计法确定相机在各个关键区域的深度面上的聚焦点具体为：根据关键区域的各个像素点的深度计算焦点堆栈的每一个像素点的梯度，对梯度图像进行均值滤波，然后计算三个通道(R通道、G通道以及B通道)梯度值的平方和取均值，再对平方和均值开根号，即得到关键区域的梯度值，焦点堆栈的同一位置像素梯度最大则说明该像素位置为该深度面的聚焦点。
步骤S302：控制相机分别对关键区域组的各个关键区域的聚焦点进行聚焦拍摄，得到各关键区域的对焦图像。
在具体应用中，在聚焦点进行拍摄则得到的图像是该深度面拍摄成像最清晰的图像，因此控制拍摄模块对该像素点(聚焦点)进行聚焦拍摄到的图像即为该关键区域的对焦图像。依次对关键区域组的各个关键区域的聚焦点进行聚焦拍摄，就能够得到该关键区域组的个关键区域的对焦图像。
实施例四：
如图4所示，在本实施例中，实施例一中的步骤S104具体包括：
步骤S401：获取各关键区域的对焦图像的模糊程度。
在具体应用中，通过拉普拉斯方差算法来各关键区域的对焦图像的模糊程度。具体的，对各关键区域的对焦图像进行灰度变换，获取各关键区域对焦图像的灰度图，对灰度图采用拉普拉斯掩模做卷积运算，再计算方差，即可得到该对焦图像的模糊程度。需要说明的是， 模糊程度是用于衡量拍摄图像的质量，模糊程度越高，则对应的图像的质量越差。在一个实施例中，上述拉普拉斯掩模为：
Figure PCTCN2019117188-appb-000001
在具体应用中，获取各个对焦图像的模糊程度也可以采用神经网络模型进行获取，将对焦图像输入到获取模糊程度的神经网络模型中，就能输出该对焦图像的模糊程度。获取模糊程度的神经网络模型可以为VGG16神经网络模型，其构建和训练过程在此不加以赘述。
步骤S402：根据各关键区域的对焦图像的模糊程度计算所述图像的模糊程度。
在具体应用中，基于关键区域的模糊程度比重，综合计算该图像的模糊程度值。计算公为：
Figure PCTCN2019117188-appb-000002
其中，Blur为所述图像的模糊程度值，Pi为第i个关键区域的比重系数，Xi为第i个关键区域的模糊程度值；为预设系数，其中P1+P2+…+Pm＝1。需要说明的是，关键区域的比重系数是根据关键区域对应的对象进行设定的，比重系数的设定可以根据经验进行设定，如包含双眼的关键区域的比重系数大于包含双眉的关键区域的比重系数。在实际应用中，预先根据划定的关键区域给定各个关键区域的比重系数，需要说明的是，关键区域的比重系数之和为1。
步骤S403：判断所述图像的模糊程度是否超过判定阈值。
步骤S404：若所述图像的模糊程度超过所述判定阈值，则确定所述检测对象为非活体。
在具体应用中，通过判断图像的模糊程度是否超过预设的判定阈值来对多个对焦图像的图像质量进行评估以判断该检测对象是否为活体，当图像的模糊程度超过所述判定阈值时，就确定该检测对象为非活体，若图像的模糊程度不超过所述判定阈值，则确定该检测对象为活体。
实施例五：
如图5所示，本实施例提供一种活体检测系统100，用于执行实施例一中的方法步骤，其包括采集模块101、估算模块102、拍摄模块103以及判断模块104。
采集模块101用于采集检测对象的图像并根据人脸的五官形状特征及分类器定位关键区域。
估算模块102用于将所述检测对象的图像的所有关键区域设置为关键区域组，估算所述关键区域组中各关键区域的各个像素点的深度。
拍摄模块103用于基于估算得到的深度分别对各关键区域进行对焦拍摄，获取各关键区域的对焦图像。
判断模块104用于根据所述各关键区域的对焦图像的图像质量判断检测对象是否为活 体。
在一个实施例中，上述人脸识别系统100还包括深度判断模块。
上述深度判断模块用于判断估算得到的各关键区域的中心点的深度是否相同，若个各关键区域的中心点的深度相同，则确定所述检测对象为非活体。
需要说明的是，本申请实施例提供的活体检测系统，由于与本申请图1所示方法实施例基于同一构思，其带来的技术效果与本申请图1所示方法实施例相同，具体内容可参见本申请图1所示方法实施例中的叙述，此处不再赘述。
因此，本实施例提供的一种活体检测系统，同样能够通过利用传统相机的对焦能力进行景深测量的方法来实现活体检测，无需增加深度摄像设备就能实现有效地活体检测，有效降低成本，解决了目前的人脸识别系统存在无法基于传统相机进行活体检测，需要增加硬件设备的问题。
实施例六：
如图6所示，在本实施例中，实施例五中的估算模块102包括用于执行图2所对应的实施例中的方法步骤的结构，其包括构建单元201、输入单元202以及确定单元203。
构建单元201用于构建和训练预估深度的神经网络模型。
输入单元202用于将采集到的图像输入到预估深度的神经网络模型中，以获取估算的所述图像的深度，图像的深度包括图像各个像素点的深度。
确定单元203用于根据各关键区域的位置及所述图像的深度确定关键区域组的各关键区域的各个像素点的深度。
实施例七：
如图7所示，在本实施例中，实施例五中的拍摄模块103包括用于执行图3所对应的实施例中的方法步骤的结构，其包括聚焦点确定单元301和拍摄单元302。
聚焦点确定单元301用于根据各关键区域的深度利用焦点堆栈深度估计法确定在各关键区域的深度面上的聚焦点。
拍摄单元302用于控制相机分别对关键区域组的各个关键区域的聚焦点进行聚焦拍摄，得到各关键区域的对焦图像。
实施例八：
如图8所示，在本实施例中，实施例五中的判断模块104包括用于执行图4所对应的实施例中的方法步骤的结构，其包括获取单元401、计算单元402以及判断单元403。
获取单元401用于获取各关键区域的对焦图像的模糊程度。
计算单元402用于根据各关键区域的对焦图像的模糊程度计算所述图像的模糊程度。
判断单元403用于判断所述图像的模糊程度是否超过判定阈值；若所述图像的模糊程度超过所述判定阈值，则确定所述检测对象为非活体。
实施例九：
图9是本申请实施例七提供的终端设备的示意图。如图9所示，该实施例的终端设备9包括：处理器90、存储器91以及存储在所述存储器91中并可在所述处理器90上运行的计算机程序92，例如程序。所述处理器90执行所述计算机程序92时实现上述各方法实施例中的步骤，例如图1所示的步骤S101至S104。或者，所述处理器90执行所述计算机程序92时实现上述系统实施例中各模块/单元的功能，例如图5所示模块101至104的功能。
示例性的，所述计算机程序92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91中，并由所述处理器90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程序指令段，该指令段用于描述所述计算机程序92在所述终端设备9中的执行过程。例如，所述计算机程序92可以被分割成采集模块、估算模块、拍摄模块以及判断模块，各模块具体功能如下：
采集模块，用于采集检测对象的图像并根据人脸的五官形状特征及分类器定位关键区域；
估算模块，用于将所述检测对象的图像的所有关键区域设置为关键区域组，估算所述关键区域组中各关键区域的各个像素点的深度；
拍摄模块，用于基于估算得到的深度分别对各关键区域进行对焦拍摄，获取各关键区域的对焦图像；
判断模块，用于根据所述各关键区域的对焦图像的图像质量判断检测对象是否为活体。
所述终端设备9可以是桌上型计算机、笔记本、掌上电脑及云端管理服务器等计算设备。所述终端设备可包括，但不仅限于，处理器90、存储器91。本领域技术人员可以理解，图9仅仅是终端设备9的示例，并不构成对终端设备9的限定，可以包括比图示更多或更少的部件，或者组合某些部件，或者不同的部件，例如所述终端设备还可以包括输入输出设备、网络接入设备、总线等。
所称处理器90可以是中央处理单元(Central Processing Unit，CPU)，还可以是其他通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现成可编程门阵列(Field-Programmable Gate Array，FPGA)或者其他可编程逻辑器件、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91可以是所述终端设备9的内部存储单元，例如终端设备9的硬盘或内存。所述存储器91也可以是所述终端设备9的外部存储设备，例如所述终端设备9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91还可以既包括所述终端设备9的内部存储单元也包括外部存储设备。所述存储器91用于存储所述计算机程序以及所述终端设备所需的其他程序和数据。所述存储器91还可以用于暂时地存储已经输出或者将要输出的数据。
所属领域的技术人员可以清楚地了解到，为了描述的方便和简洁，仅以上述各功能单 元、模块的划分进行举例说明，实际应用中，可以根据需要而将上述功能分配由不同的功能单元、模块完成，即将所述系统的内部结构划分成不同的功能单元或模块，以完成以上描述的全部或者部分功能。实施例中的各功能单元、模块可以集成在一个处理单元中，也可以是各单元单独物理存在，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。另外，各功能单元、模块的具体名称也只是为了便于相互区分，并不用于限制本申请的保护范围。上述无线终端中单元、模块的具体工作过程，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。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
在本申请所提供的实施例中，应该理解到，所揭露的系统/终端设备和方法，可以通过其它的方式实现。例如，以上所描述的系统/终端设备实施例仅仅是示意性的，例如，所述模块或单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通讯连接可以是通过一些接口，系统或单元的间接耦合或通讯连接，可以是电性，机械或其它的形式。
所述设置为分离部件说明的单元可以是或者也可以不是物理上分开的，设置为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各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。
所述集成的模块/单元如果以软件功能单元的形式实现并设置为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本申请实现上述实施例方法中的全部或部分流程，也可以通过计算机程序来指令相关的硬件来完成，所述的计算机程序可存储于一计算机可读存储介质中，该计算机程序在被处理器执行时，可实现上述各方法实施例的步骤。其中，所述计算机程序包括计算机程序代码，所述计算机程序代码可以为源代码形式、对象代码形式、可执行文件或某些中间形式等。所述计算机可读介质 可以包括：能够携带所述计算机程序代码的任何实体或系统、记录介质、U盘、移动硬盘、磁碟、光盘、计算机存储器、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、电载波信号、电信信号以及软件分发介质等。需要说明的是，所述计算机可读介质包含的内容可以根据司法管辖区内立法和专利实践的要求进行适当的增减，例如在某些司法管辖区，根据立法和专利实践，计算机可读介质不包括是电载波信号和电信信号。
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，均应包含在本申请的保护范围之内。

Claims (20)

  1. 一种活体检测方法，其特征在于，包括：
    采集检测对象的图像并根据人脸的五官形状特征及分类器定位关键区域；
    将所述检测对象的图像的所有关键区域设置为关键区域组，估算所述关键区域组中各关键区域的各个像素点的深度；
    基于估算得到的深度分别对各关键区域进行对焦拍摄，获取各关键区域的对焦图像；
    根据所述各关键区域的对焦图像的图像质量判断检测对象是否为活体。
  2. 根据权利要求1所述的活体检测方法，其特征在于，在设置关键区域组，估算所述关键区域组中各关键区域的深度之后，还包括：
    判断估算得到的各关键区域的中心点的深度是否相同；
    若个各关键区域的中心点的深度相同，则确定所述检测对象为非活体。
  3. 根据权利要求1所述的活体检测方法，其特征在于，所述将所述检测对象的图像的所有关键区域设置为关键区域组，估算所述关键区域组中各关键区域的各个像素点的深度，包括：
    构建和训练预估深度的神经网络模型；
    将采集到的图像输入到预估深度信息的神经网络模型中，以获取估算的所述图像的深度，图像的深度包括图像各个像素点的深度；
    根据各关键区域的位置及所述图像的深度确定关键区域组的各关键区域的各个像素点的深度。
  4. 根据权利要求1所述的活体检测方法，其特征在于，基于估算得到的深度分别对各关键区域进行对焦拍摄，获取各关键区域的对焦图像，包括：
    根据各关键区域的深度利用焦点堆栈深度估计法确定在各关键区域的深度面上的聚焦点；
    控制相机分别对关键区域组的各个关键区域的聚焦点进行聚焦拍摄，得到各关键区域的对焦图像。
  5. 根据权利要求1所述的活体检测方法，其特征在于，所述根据所述各关键区域的对焦图像的图像质量判断检测对象是否为活体，包括：
    获取各关键区域的对焦图像的模糊程度；
    根据各关键区域的对焦图像的模糊程度计算所述图像的模糊程度；
    判断所述图像的模糊程度是否超过判定阈值；
    若所述图像的模糊程度超过所述判定阈值，则确定所述检测对象为非活体。
  6. 一种活体检测系统，其特征在于，包括：
    采集模块，用于采集检测对象的图像并根据人脸的五官形状特征及分类器定位关键区 域；
    估算模块，用于将所述检测对象的图像的所有关键区域设置为关键区域组，估算所述关键区域组中各关键区域的各个像素点的深度；
    拍摄模块，用于基于估算得到的深度分别对各关键区域进行对焦拍摄，获取各关键区域的对焦图像；
    判断模块，用于根据所述各关键区域的对焦图像的图像质量判断检测对象是否为活体。
  7. 根据权利要求6所述的活体检测系统，其特征在于，所述拍摄模块包括：
    聚焦点确定单元，用于根据各关键区域的深度利用焦点堆栈深度估计法确定在各关键区域的深度面上的聚焦点；
    拍摄单元，用于控制相机分别对关键区域组的各个关键区域的聚焦点进行聚焦拍摄，得到各关键区域的对焦图像。
  8. 根据权利要求6所述的活体检测系统，其特征在于，所述判断模块包括：
    获取单元，用于获取各关键区域的对焦图像的模糊程度；
    计算单元，用于根据各关键区域的对焦图像的模糊程度计算所述图像的模糊程度；
    判断单元，用于判断所述图像的模糊程度是否超过判定阈值；若所述图像的模糊程度超过所述判定阈值，则确定所述检测对象为非活体。
  9. 根据权利要求6所述的活体检测系统，其特征在于，还包括：
    深度判断模块，用于判断估算得到的各关键区域的中心点的深度是否相同；若个各关键区域的中心点的深度相同，则确定所述检测对象为非活体。
  10. 根据权利要求6所述的活体检测系统，其特征在于，所述估算模块包括：
    构建单元，用于构建和训练预估深度的神经网络模型；
    输入单元，用于将采集到的图像输入到预估深度的神经网络模型中，以获取估算的所述图像的深度，图像的深度包括图像各个像素点的深度；
    确定单元，用于根据各关键区域的位置及所述图像的深度确定关键区域组的各关键区域的各个像素点的深度。
  11. 一种终端设备，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，其特征在于，所述处理器执行所述计算机可读指令时实现如下步骤：
    采集检测对象的图像并根据人脸的五官形状特征及分类器定位关键区域；
    将所述检测对象的图像的所有关键区域设置为关键区域组，估算所述关键区域组中各关键区域的各个像素点的深度；
    基于估算得到的深度分别对各关键区域进行对焦拍摄，获取各关键区域的对焦图像；
    根据所述各关键区域的对焦图像的图像质量判断检测对象是否为活体。
  12. 根据权利要求11所述的终端设备，其特征在于，在设置关键区域组，估算所述关 键区域组中各关键区域的深度之后，还包括：
    判断估算得到的各关键区域的中心点的深度是否相同；
    若个各关键区域的中心点的深度相同，则确定所述检测对象为非活体。
  13. 根据权利要求11所述的终端设备，其特征在于，所述将所述检测对象的图像的所有关键区域设置为关键区域组，估算所述关键区域组中各关键区域的各个像素点的深度，包括：
    构建和训练预估深度的神经网络模型；
    将采集到的图像输入到预估深度信息的神经网络模型中，以获取估算的所述图像的深度，图像的深度包括图像各个像素点的深度；
    根据各关键区域的位置及所述图像的深度确定关键区域组的各关键区域的各个像素点的深度。
  14. 根据权利要求11所述的终端设备，其特征在于，基于估算得到的深度分别对各关键区域进行对焦拍摄，获取各关键区域的对焦图像，包括：
    根据各关键区域的深度利用焦点堆栈深度估计法确定在各关键区域的深度面上的聚焦点；
    控制相机分别对关键区域组的各个关键区域的聚焦点进行聚焦拍摄，得到各关键区域的对焦图像。
  15. 根据权利要求11所述的终端设备，其特征在于，所述根据所述各关键区域的对焦图像的图像质量判断检测对象是否为活体，包括：
    获取各关键区域的对焦图像的模糊程度；
    根据各关键区域的对焦图像的模糊程度计算所述图像的模糊程度；
    判断所述图像的模糊程度是否超过判定阈值；
    若所述图像的模糊程度超过所述判定阈值，则确定所述检测对象为非活体。
  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：
    采集检测对象的图像并根据人脸的五官形状特征及分类器定位关键区域；
    将所述检测对象的图像的所有关键区域设置为关键区域组，估算所述关键区域组中各关键区域的各个像素点的深度；
    基于估算得到的深度分别对各关键区域进行对焦拍摄，获取各关键区域的对焦图像；
    根据所述各关键区域的对焦图像的图像质量判断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，在设置关键区域组，估算所述关键区域组中各关键区域的深度之后，还包括：
    判断估算得到的各关键区域的中心点的深度是否相同；
    若个各关键区域的中心点的深度相同，则确定所述检测对象为非活体。
  18. 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，所述将所述检测对象的图像的所有关键区域设置为关键区域组，估算所述关键区域组中各关键区域的各个像素点的深度，包括：
    构建和训练预估深度的神经网络模型；
    将采集到的图像输入到预估深度信息的神经网络模型中，以获取估算的所述图像的深度，图像的深度包括图像各个像素点的深度；
    根据各关键区域的位置及所述图像的深度确定关键区域组的各关键区域的各个像素点的深度。
  19. 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，基于估算得到的深度分别对各关键区域进行对焦拍摄，获取各关键区域的对焦图像，包括：
    根据各关键区域的深度利用焦点堆栈深度估计法确定在各关键区域的深度面上的聚焦点；
    控制相机分别对关键区域组的各个关键区域的聚焦点进行聚焦拍摄，得到各关键区域的对焦图像。
  20. 根据权利要求16所述的计算机非易失性可读存储介质，其特征在于所述根据所述各关键区域的对焦图像的图像质量判断检测对象是否为活体，包括：
    获取各关键区域的对焦图像的模糊程度；
    根据各关键区域的对焦图像的模糊程度计算所述图像的模糊程度；
    判断所述图像的模糊程度是否超过判定阈值；
    若所述图像的模糊程度超过所述判定阈值，则确定所述检测对象为非活体。
PCT/CN2019/117188 2019-02-13 2019-11-11 一种活体检测方法、系统及终端设备 WO2020164266A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910112917.3A CN109948439B (zh) 2019-02-13 2019-02-13 一种活体检测方法、系统及终端设备
CN201910112917.3 2019-02-13

Publications (1)

Publication Number Publication Date
WO2020164266A1 true WO2020164266A1 (zh) 2020-08-20

Family

ID=67007934

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/117188 WO2020164266A1 (zh) 2019-02-13 2019-11-11 一种活体检测方法、系统及终端设备

Country Status (2)

Country Link
CN (1) CN109948439B (zh)
WO (1) WO2020164266A1 (zh)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109948439B (zh) * 2019-02-13 2023-10-31 平安科技（深圳）有限公司 一种活体检测方法、系统及终端设备
CN110929557A (zh) * 2019-09-25 2020-03-27 四川大学锦城学院 一种基于活体检测的智能安防方法、系统及处理装置
CN111241989B (zh) * 2020-01-08 2023-06-13 腾讯科技（深圳）有限公司 图像识别方法及装置、电子设备
CN111814697B (zh) * 2020-07-13 2024-02-13 伊沃人工智能技术（江苏）有限公司 一种实时人脸识别方法、系统及电子设备
CN113903084A (zh) * 2021-10-13 2022-01-07 深圳前海微众银行股份有限公司 活体检测方法、设备、可读存储介质及计算机程序产品

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150016676A1 (en) * 2013-07-10 2015-01-15 Soongsil University Research Consortium Techno-Park System and method for detecting object using depth information
CN105023010A (zh) * 2015-08-17 2015-11-04 中国科学院半导体研究所 一种人脸活体检测方法及系统
CN105335722A (zh) * 2015-10-30 2016-02-17 商汤集团有限公司 一种基于深度图像信息的检测系统及方法
CN109034102A (zh) * 2018-08-14 2018-12-18 腾讯科技（深圳）有限公司 人脸活体检测方法、装置、设备及存储介质
CN109117755A (zh) * 2018-07-25 2019-01-01 北京飞搜科技有限公司 一种人脸活体检测方法、系统和设备
CN109948439A (zh) * 2019-02-13 2019-06-28 平安科技（深圳）有限公司 一种活体检测方法、系统及终端设备

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5004726B2 (ja) * 2007-09-05 2012-08-22 キヤノン株式会社 撮像装置、レンズユニットおよび制御方法
US9036936B2 (en) * 2011-06-20 2015-05-19 Fujifilm Corporation Image processing device, image processing method, and image processing program
US9025874B2 (en) * 2013-02-19 2015-05-05 Blackberry Limited Method and system for generating shallow depth of field effect
CN105975926B (zh) * 2016-04-29 2019-06-21 中山大学 基于光场相机的人脸活体检测方法
CN107016348B (zh) * 2017-03-09 2022-11-22 Oppo广东移动通信有限公司 结合深度信息的人脸检测方法、检测装置和电子装置

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150016676A1 (en) * 2013-07-10 2015-01-15 Soongsil University Research Consortium Techno-Park System and method for detecting object using depth information
CN105023010A (zh) * 2015-08-17 2015-11-04 中国科学院半导体研究所 一种人脸活体检测方法及系统
CN105335722A (zh) * 2015-10-30 2016-02-17 商汤集团有限公司 一种基于深度图像信息的检测系统及方法
CN109117755A (zh) * 2018-07-25 2019-01-01 北京飞搜科技有限公司 一种人脸活体检测方法、系统和设备
CN109034102A (zh) * 2018-08-14 2018-12-18 腾讯科技（深圳）有限公司 人脸活体检测方法、装置、设备及存储介质
CN109948439A (zh) * 2019-02-13 2019-06-28 平安科技（深圳）有限公司 一种活体检测方法、系统及终端设备

Also Published As

Publication number Publication date
CN109948439A (zh) 2019-06-28
CN109948439B (zh) 2023-10-31

Similar Documents

Publication Publication Date Title
WO2020164266A1 (zh) 一种活体检测方法、系统及终端设备
CN110826519B (zh) 人脸遮挡检测方法、装置、计算机设备及存储介质
CN111091075B (zh) 人脸识别方法、装置、电子设备及存储介质
WO2019153504A1 (zh) 一种群组创建的方法及其终端
CN111368717B (zh) 视线确定方法、装置、电子设备和计算机可读存储介质
TW202026948A (zh) 活體檢測方法、裝置以及儲存介質
CN105243371A (zh) 一种人脸美颜程度的检测方法、系统及拍摄终端
CN107622497B (zh) 图像裁剪方法、装置、计算机可读存储介质和计算机设备
WO2021008205A1 (zh) 图像处理
CN111008935B (zh) 一种人脸图像增强方法、装置、系统及存储介质
CN111127303A (zh) 背景虚化方法、装置、终端设备及计算机可读存储介质
TWI731503B (zh) 活體臉部辨識系統與方法
CN110147708A (zh) 一种图像数据处理方法和相关装置
WO2020143165A1 (zh) 一种翻拍图像的识别方法、系统及终端设备
CN111814564A (zh) 基于多光谱图像的活体检测方法、装置、设备和存储介质
CN111444555B (zh) 一种测温信息显示方法、装置及终端设备
WO2023142352A1 (zh) 一种深度图像的获取方法、装置、终端、成像系统和介质
CN107368817B (zh) 人脸识别方法和装置
CN113298158A (zh) 数据检测方法、装置、设备及存储介质
CN112802081A (zh) 一种深度检测方法、装置、电子设备及存储介质
WO2022199395A1 (zh) 人脸活体检测方法、终端设备及计算机可读存储介质
CN110502961B (zh) 一种面部图像检测方法及装置
CN111126250A (zh) 一种基于ptgan的行人重识别方法及装置
CN109726613B (zh) 一种用于检测的方法和装置
CN115049819A (zh) 注视区域识别方法及装置

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: 19915023

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19915023

Country of ref document: EP

Kind code of ref document: A1