WO2020083183A1 - 活体检测方法、装置、电子设备、存储介质及应用活体检测方法的相关系统 - Google Patents

活体检测方法、装置、电子设备、存储介质及应用活体检测方法的相关系统 Download PDF

Info

Publication number
WO2020083183A1
WO2020083183A1 PCT/CN2019/112196 CN2019112196W WO2020083183A1 WO 2020083183 A1 WO2020083183 A1 WO 2020083183A1 CN 2019112196 W CN2019112196 W CN 2019112196W WO 2020083183 A1 WO2020083183 A1 WO 2020083183A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
current image
feature
detected
feature vector
Prior art date
Application number
PCT/CN2019/112196
Other languages
English (en)
French (fr)
Inventor
罗文寒
王耀东
刘威
Original Assignee
腾讯科技（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 腾讯科技（深圳）有限公司 filed Critical 腾讯科技（深圳）有限公司
Priority to EP19875568.8A priority Critical patent/EP3761222B1/en
Publication of WO2020083183A1 publication Critical patent/WO2020083183A1/zh
Priority to US17/070,435 priority patent/US11721087B2/en
Priority to US18/333,363 priority patent/US20230343071A1/en
Priority to US18/333,357 priority patent/US20230343070A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/40—Spoof detection, e.g. liveness detection
    • G06V40/45—Detection of the body part being alive
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/161—Detection; Localisation; Normalisation
    • G06V40/165—Detection; Localisation; Normalisation using facial parts and geometric relationships
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/168—Feature extraction; Face representation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/172—Classification, e.g. identification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • 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
    • G06V40/18—Eye characteristics, e.g. of the iris
    • G06V40/193—Preprocessing; Feature extraction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/40—Spoof detection, e.g. liveness detection

Definitions

  • the embodiments of the present application relate to the technical field of facial feature recognition, and in particular, to a living body detection method, device, electronic device, storage medium, and payment system, video monitoring system, and access control system applying the living body detection method.
  • biometric recognition is widely used, for example, face payment, face recognition in video surveillance, and fingerprint recognition and iris recognition in access control authorization.
  • biometrics there are various threats to biometrics, such as attackers using forged faces, fingerprints, irises, etc. for biometrics.
  • Various embodiments of the present application provide a living body detection method, device, electronic equipment, storage medium, and payment system, video monitoring system, and access control system applying the living body detection method.
  • An embodiment of the present application provides a living body detection method, which is executed by an electronic device and includes: traversing a plurality of images of an object to be detected, and using the currently traversed image as a current image; performing facial feature extraction on the current image to obtain A feature vector corresponding to the current image, where the feature vector is used to describe the structure of the facial features of the object to be detected in the current image; the feature vector corresponding to the current image corresponds to the historical image in the feature sequence A change of a feature vector to capture the action behavior of the object to be detected, the historical image is an image that has been traversed among the multiple images, and the feature sequence includes a feature vector corresponding to at least one historical image; if the captured If the object to be detected has an action behavior, it is determined that the object to be detected is a living body.
  • An embodiment of the present application provides a living body detection device, including: an image traversal module for traversing a plurality of images of an object to be detected, using the currently traversed image as a current image; and a feature extraction module for detecting the current image Perform facial feature extraction to obtain a feature vector corresponding to the current image, where the feature vector is used to describe the structure of the facial features of the object to be detected in the current image; a behavior capture module is used to determine The change of the feature vector corresponding to the image relative to the feature vector corresponding to the historical image in the feature sequence to capture the action behavior of the object to be detected, the historical image is an image that has been traversed among the plurality of images, and the feature sequence includes at least A feature vector corresponding to a historical image; a living body detection module, configured to determine that the object to be detected is a living body if the captured object has an action behavior.
  • An embodiment of the present application provides an electronic device, including a processor and a memory, where computer-readable instructions are stored on the memory, and when the computer-readable instructions are executed by the processor, the living body detection method described above is implemented.
  • An embodiment of the present application provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the living body detection method as described above is implemented.
  • An embodiment of the present application provides a payment system, the payment system includes a payment terminal and a payment server, wherein the payment terminal is used to collect multiple images of a payment user; the payment terminal includes a living body detection device used to A plurality of images of the payment user determine corresponding feature vectors, and capture the action behavior of the payment user according to the relative change between the determined feature vectors, and if the action behavior of the payment user is captured, the payment is determined
  • the user is a living body; when the payment user is a living body, the payment terminal authenticates the payment user to initiate a payment request to the payment server when the payment user passes the identity verification.
  • An embodiment of the present application provides a video monitoring system.
  • the video monitoring system includes a monitoring screen, a plurality of cameras, and a monitoring server, wherein a plurality of the cameras are used to collect multiple images of a monitored object; the monitoring server includes a living body detection
  • the device is used to determine the corresponding feature vector according to the multiple images of the monitored object, and to capture the action behavior of the monitored object according to the relative change between the determined feature vectors, if the action behavior of the monitored object is captured, It is determined that the monitoring object is a living body; when the monitoring object is a living body, the monitoring server identifies the monitoring object to obtain a tracking target, and the tracking target is displayed on the monitoring screen through an image screen Carry out video surveillance.
  • An embodiment of the present application provides an access control system
  • the access control system includes a reception device, an identification server, and an access control device, wherein the reception device is used to collect multiple images of an in and out object;
  • the identification server includes a living body detection device , Used to determine the corresponding feature vector according to the multiple images of the access object, and to capture the action behavior of the access object according to the relative change between the determined feature vectors, if the action behavior of the access object is captured, It is determined that the access object is a living body; when the access object is a living body, the identification server performs identification on the access object, so that the access control device configures an access authority for the access object that successfully completes the identification, so that the The access object controls the access gate of the designated work area to perform the release action according to the configured access control authority.
  • Fig. 1 is a block diagram of a hardware structure of an electronic device according to an exemplary embodiment
  • Fig. 2 is a flowchart of a method for detecting a living body according to an exemplary embodiment
  • Fig. 3 is a flowchart of another method for detecting a living body according to an exemplary embodiment
  • FIG. 4 is a flowchart of step 330 in one embodiment in the embodiment corresponding to FIG. 2;
  • FIG. 5 is a schematic diagram of several key points of eyes in the image involved in the embodiment corresponding to FIG. 4;
  • FIG. 6 is a schematic diagram of the trend of the aspect ratio of the eye involved in the embodiment corresponding to FIG. 4;
  • step 7 is a flowchart of step 331 in an embodiment corresponding to the embodiment of FIG. 4;
  • FIG. 8 is a schematic diagram of the index relationship constructed by the face key point model involved in the embodiment corresponding to FIG. 7;
  • step 9 is a flowchart of step 430 in an embodiment corresponding to the embodiment of FIG. 3;
  • FIG. 10 is a schematic diagram of a specific implementation of performing a queue-in operation / queue-out operation for a feature vector corresponding to an image in a queue involved in the embodiment corresponding to FIG. 9;
  • FIG. 11 is a flowchart of step 350 in one embodiment in the embodiment corresponding to FIG. 2;
  • FIG. 12 is a flowchart of step 355 in an embodiment corresponding to FIG. 11;
  • step 370 is a flowchart of step 370 in an embodiment corresponding to FIG. 2 in an embodiment
  • FIG. 14 is a schematic diagram of an implementation environment based on identity verification in an application scenario
  • 15 is a schematic diagram of an implementation environment based on identity recognition in an application scenario
  • 16 is a schematic diagram of another implementation environment based on identity recognition in an application scenario
  • 17 is a specific sequence diagram of a living body detection method in an application scenario
  • FIG. 18 is a schematic diagram of a specific implementation of the living body detection method involved in the application scenario of FIG. 17;
  • Fig. 19 is a block diagram of a living body detection device according to an exemplary embodiment
  • Fig. 20 is a block diagram of an electronic device according to an exemplary embodiment.
  • the embodiment of the present application takes the biological feature as the facial feature as an example for description.
  • the living body detection method is to perform living body detection on the image of the object to be detected, that is, to detect whether the facial feature contour of the object to be detected has changed in the image, if it is detected that the facial feature contour of the object to be detected has occurred in the image Changes, that is, the object to be detected is determined to be a living body.
  • the facial features of the object to be detected in the image are eyes or mouth.
  • the object to be detected blinks or opens its mouth, it will cause the contour of the facial features in the image to change, and thus it can be determined that the object to be detected is a living body.
  • the prosthesis attack sample refers to the image of the object to be detected stolen by the attacker. It is the attacker's use of the characteristics of the eyes or mouth to tamper with the image of the object to be detected. The outline of the middle eye, or the outline of the mouth in the image is blocked with a pen, so that the contour of the facial features in the image has changed, causing the prosthesis (camouflaged object to be detected, that is, the attacker) to blink, open the mouth and other artifacts The prosthesis was misjudged as a living body.
  • the living body detection method is based on a piece of video of the object to be detected, since the living body detection is performed in units of image frames, the attacker can still rely on the prosthesis attack sample to easily crack the above living body detection method, for example, quickly blocking multiple consecutive times The contour of facial features in the image, which makes the prosthesis misjudged as a living body.
  • the embodiments of the present application specifically propose a living body detection method, which can effectively improve the defense against the prosthesis attack samples and has high security.
  • This kind of living body detection method is realized by a computer program, and correspondingly, the constructed living body detection device can be stored in an electronic device with a von Neumann system architecture to be executed in the electronic device, thereby realizing the object to be detected Biopsy.
  • the electronic device may be a smart phone, tablet computer, notebook computer, desktop computer, server, etc., which is not limited herein.
  • FIG. 1 is a block diagram of an electronic device according to an exemplary embodiment of the present application. It should be noted that this type of electronic device is only an example adapted to the embodiments of the present application, and cannot be considered as providing any limitation on the scope of use of the embodiments of the present application. This type of electronic device cannot also be interpreted as requiring or having to depend on one or more components in the exemplary electronic device 100 shown in FIG. 1.
  • the electronic device 100 includes a memory 101, a storage controller 103, one or more (only one is shown in FIG. 1) processor 105, a peripheral interface 107, a radio frequency module 109, a positioning module 111, a camera module 113, an audio module 115, a touch screen 117, and a key module 119. These components communicate with each other through one or more communication buses / signal lines 121.
  • the memory 101 may be used to store computer programs and modules, such as the computer programs and modules corresponding to the living body detection method and apparatus in the exemplary embodiments of the present application, and the processor 105 executes the computer program stored in the memory 101 to execute Various functions and data processing, for example, to complete the living body detection method described in any embodiment of the present application.
  • the memory 101 serves as a carrier for resource storage, and may be a random access memory, such as a high-speed random access memory, a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other solid-state memory.
  • the storage method can be short-term storage or permanent storage.
  • the peripheral interface 107 may include at least one wired or wireless network interface, at least one serial-to-parallel conversion interface, at least one input-output interface, and at least one USB interface, etc., for coupling various external input / output devices to the memory 101 and the processor 105, to achieve communication with various external input / output devices.
  • the radio frequency module 109 is used to send and receive electromagnetic waves to realize the mutual conversion of electromagnetic waves and electrical signals, so as to communicate with other devices through the communication network.
  • the communication network includes a cellular telephone network, a wireless local area network, or a metropolitan area network.
  • the above communication network can use various communication standards, protocols, and technologies.
  • the positioning module 111 is used to obtain the current geographic location of the electronic device 100.
  • Examples of the positioning module 111 include, but are not limited to, global satellite positioning system (GPS), positioning technology based on a wireless local area network or a mobile communication network.
  • GPS global satellite positioning system
  • the camera module 113 belongs to the camera and is used to take pictures or videos.
  • the captured pictures or videos can be stored in the memory 101, and can also be sent to the upper computer through the radio frequency module 109.
  • the camera module 113 is used to photograph the object to be detected to form an image of the object to be detected.
  • the audio module 115 provides an audio interface to the user, which may include one or more microphone interfaces, one or more speaker interfaces, and one or more headphone interfaces. Perform audio data interaction with other devices through the audio interface.
  • the audio data may be stored in the memory 101, and may also be sent through the radio frequency module 109.
  • the touch screen 117 provides an input and output interface between the electronic device 100 and the user. Specifically, the user may perform input operations through the touch screen 117, such as gesture operations such as tap, touch, and slide, so that the electronic device 100 responds to the input operations.
  • the electronic device 100 displays and outputs the output content formed in any form or combination of text, pictures or videos to the user through the touch screen 117.
  • the key module 119 includes at least one key to provide an interface for the user to input to the electronic device 100.
  • the user can press the different keys to cause the electronic device 100 to perform different functions.
  • the sound adjustment button can be used by the user to adjust the volume of the sound played by the electronic device 100.
  • FIG. 1 is merely an illustration, and the electronic device 100 may further include more or fewer components than those shown in FIG. 1 or have different components than those shown in FIG. 1.
  • Each component shown in FIG. 1 may be implemented by hardware, software, or a combination thereof.
  • a living body detection method is applicable to an electronic device.
  • the structure of the electronic device may be as shown in FIG. 1.
  • This living body detection method can be executed by an electronic device, and can include the following steps:
  • Step 310 Traverse multiple images to be detected, and use the currently traversed image as the current image.
  • multiple images of the object to be detected may refer to a piece of video, that is, the camera device generates a single image of the object to be detected; it may also refer to multiple photos, that is, the camera device takes multiple consecutive photos of the object to be detected Generated.
  • the living body detection may be performed based on a piece of video of the object to be detected, or may be performed based on multiple photos of the object to be detected, which is not limited in this embodiment.
  • the acquisition of multiple images may be multiple images acquired in real time, or multiple images stored in advance, that is, multiple images acquired by reading a historical time period in the cache area, this embodiment also There is no restriction on this.
  • live detection can be performed on the multiple images of the object to be detected in real time, or the image of the object to be detected can be stored first and processed later. For example, when the electronic device has less processing tasks, or according to the instructions of the inspector.
  • the camera device may be a video camera, a video recorder, or other electronic devices with image acquisition functions, such as smart phones.
  • the living body detection is performed in units of image frames. Therefore, after acquiring multiple images of the object to be detected, multiple images of the object to be detected can be traversed, so as to facilitate living body detection according to the traversed images.
  • the currently traversed image is regarded as the current image
  • the traversed image is regarded as the historical image.
  • Step 330 Perform facial feature extraction on the current image to obtain a feature vector corresponding to the current image, where the feature vector is used to describe a face feature structure of the object to be detected in the current image.
  • the facial features of the object to be detected may be, for example, eyes, mouth, ears, iris, etc. For this reason, corresponding to the feature vector of the image, it is used to achieve an accurate description of the structure of the facial features of the object to be detected in the image, and to uniquely represent the structure of the facial features of the object to be detected in the image.
  • the structure of the facial features in the image will be different, so that the feature vectors corresponding to the image are also different.
  • the feature vector corresponding to the image is the aspect ratio of the eye to describe the structure of the eye of the object to be detected in the image;
  • the facial feature of the object to be detected is the mouth, the image corresponds to
  • the feature vector of is the aspect ratio of the mouth to describe the structure of the mouth of the object to be detected in the image.
  • the feature vectors corresponding to the images are not listed one by one. Different objects to be detected have their corresponding images, and then there are feature vectors corresponding to the corresponding images, so as to accurately describe the different objects to be detected.
  • the structure of facial features in the corresponding image is not listed one by one.
  • Step 350 Capture the action behavior of the object to be detected according to the change of the feature vector corresponding to the current image relative to the feature vector corresponding to the historical image in the feature sequence.
  • the currently traversed image is regarded as the current image
  • the historical image refers to the traversed image among multiple images.
  • the feature sequence is generated by tracking the facial features of the object to be detected in multiple images, and the feature sequence includes at least one feature vector corresponding to the historical image.
  • the construction process of the feature sequence may include the following steps:
  • Step 410 Compare the feature vector corresponding to the current image with the normal structure interval.
  • Step 430 If the feature vector corresponding to the current image is within the normal structure interval, add the feature vector corresponding to the current image to the feature sequence.
  • Step 450 If the feature vector corresponding to the current image is outside the normal structure interval, ignore the feature vector corresponding to the current image and continue to traverse the next image in the multiple images.
  • the structure of the facial features is relatively fixed. Therefore, the structure of the face features of the object to be detected in the image is also relatively fixed, which is regarded as a normal structure. For example, if the facial feature of the object to be detected is an eye, the outline of the eye when the eye is opened is regarded as a normal structure.
  • the normal structure interval represents the fluctuation range of the normal structure of the facial features of the object to be detected in the image.
  • This normal structure interval can be flexibly set according to the actual requirements of the application scenario. For example, in an application scenario with high accuracy requirements, a normal structure interval with a narrow fluctuation range is set, which is not limited in this embodiment.
  • the feature sequence essentially reflects the normal structure of the facial features of the object to be detected in the historical image. It can also be understood that the feature sequence is used to achieve an accurate description of the normal structure of the facial features of the object to be detected during the historical image acquisition period.
  • the feature vector corresponding to the current image changes from the feature vector corresponding to the historical image in the feature sequence, it indicates that the structure of the facial feature of the object to be detected in the current image, and the object to be detected described by the relative feature sequence is in the history
  • the normal structure of facial features has changed during the image acquisition period.
  • the face feature of the object to be detected is still described as an eye.
  • the normal structure is the outline of the eye when the eye is opened.
  • the changed structure is the outline of the eye when blinking.
  • the action behavior of the object to be detected includes, but is not limited to: blinking behavior, opening mouth behavior, closing mouth behavior, beckoning behavior, stomping behavior, and so on.
  • the object to be detected when the feature vector corresponding to one image (current image) has changed relative to the feature vector corresponding to the previous images (historical images), it indicates that the object to be detected is
  • the facial feature contours in the image have changed, for example, when the object to be detected blinks, at this time, it is regarded as capturing the action behavior of the object to be detected, and then the object to be detected is determined to be a living body.
  • Step 370 if it is captured that the object to be detected has an action behavior, it is determined that the object to be detected is a living body.
  • the blinking behavior, opening mouth behavior, closing mouth behavior, beckoning behavior, or stomping behavior of the object to be detected is captured, it can be determined that the object to be detected is a living body.
  • a living body detection scheme based on the relative change of the feature vector is realized, that is, for multiple images of the object to be detected, only the feature vector corresponding to one image occurs relative to the feature vector corresponding to the previous several images Relative changes, the object to be detected will be judged as a living body, so as to filter the false judgment of the prosthesis caused by the mutation of the facial feature contour in the prosthesis attack sample, thereby effectively improving the defensiveness of the living body detection method against the prosthesis attack sample , With high security.
  • step 330 may include the following steps:
  • Step 331 Identify the facial features in the current image to obtain several key points of the facial features in the current image.
  • the feature vector is used to describe the structure of the facial features of the object to be detected in the image
  • the structure of the facial features in the image essentially refers to the outline of the facial features in the image.
  • the structure of the eye of the object to be detected in the image refers to the outline of the eye in the image.
  • the contour of the facial features in the image can be regarded as consisting of a series of pixels.
  • Facial feature recognition it is not necessary to use all the pixels that constitute the contour of the facial features in the image. Therefore, in this embodiment, by Facial feature recognition, to obtain some key points of facial features in the image.
  • the six key points of the eyes in the image include: one key pixel point P1 of the right eye corner, two key pixel points P2 of the upper eyelid And P3, one key pixel point P4 of the left eye corner, and two key pixel points P5 and P6 of the lower eyelid.
  • facial feature recognition may be implemented in different ways according to different types of objects to be detected.
  • facial feature recognition may be implemented using a face key point model, that is, a face key point model is called to extract several key points of facial features in the image.
  • Step 333 Calculate the structural distance ratio of the facial features according to several key points of the facial features in the current image to obtain the feature vector corresponding to the current image.
  • the feature vector corresponding to the image is the aspect ratio of the eye to describe the structure of the eye of the object to be detected in the image; if the facial feature of the object to be detected is the mouth Then, the feature vector corresponding to the image is the aspect ratio of the mouth to describe the structure of the mouth of the object to be detected in the image.
  • the aspect ratio of the eyes represents the proportion of the structural distance of the eyes
  • the aspect ratio of the mouth represents the proportion of the structural distance of the mouth
  • the outline of the eyes in the image can be basically located, thereby reflecting the structure of the eyes in the image.
  • the calculation formula of the eye aspect ratio is shown in (1):
  • EAR is the aspect ratio of the eye
  • p 1 is the coordinate of the key pixel at the right eye corner
  • p 2 and p 3 are the coordinates of the two key pixel at the upper eyelid
  • p 4 is the coordinate of the key pixel at the left eye corner
  • p 5 and p 6 represent the coordinates of the two key pixels where the lower eyelid is located.
  • represents the norm of the coordinate difference between a pair of key pixel points where the left and right eye corners are located
  • represents a pair of key pixel points where the upper and lower eyelids are located
  • represents the norm of the coordinate difference between another pair of key pixel points where the upper and lower eyelids are located.
  • the numerator represents the vertical distance between the upper and lower eyelids of the eye
  • the denominator represents the horizontal distance between the left and right corners of the eye. It should be noted that since the numerator contains two sets of vertical distances, and the denominator contains only a set of horizontal distances, the denominator is weighted, that is ⁇ 2.
  • an accurate description of the facial feature structure in the image is realized, which provides a basis for subsequently capturing whether the object to be detected has an action behavior, thereby enabling live detection.
  • the facial features in the current image are human facial features.
  • facial features include, but are not limited to: eyebrows, eyes, nose, mouth, ears, and so on.
  • step 331 may include the following steps:
  • Step 3311 Perform grayscale processing on the current image to obtain a grayscale image of the current image.
  • Step 3313 Input the grayscale image of the current image into a key point model of a human face to perform facial feature recognition, and obtain several key points of facial features in the current image.
  • the face key point model essentially constructs an index relationship for the face features in the image, so that the key points of a specific face feature can be extracted from the image through the constructed index relationship.
  • the indexes of the six key points of the left and right eyes are 37 to 42 and 43 to 48, respectively, and the indexes of the twenty key points of the mouth are 49 to 68.
  • the coordinates of several key points of the face features that have been index-marked in the image are stored accordingly, thereby constructing the index relationship between the index and the coordinates for the face features in the image.
  • the coordinates of several key points of facial features in the image can be obtained from the index.
  • the face key point model is generated by performing model training on a specified mathematical model through a large number of image samples.
  • the image sample refers to an image that has been indexed.
  • Model training is essentially to iteratively optimize the parameters of the specified mathematical model so that the specified algorithm function constructed from the parameters satisfies the convergence conditions.
  • specified mathematical models including but not limited to: machine learning models such as logistic regression, support vector machines, random forests, neural networks, etc.
  • Specify algorithm functions including but not limited to: maximum expectation function, loss function, etc.
  • the parameters of the specified mathematical model are updated, and the loss value of the loss function constructed by the updated parameters is calculated according to the latter image sample.
  • the feature sequence is a queue of a specified length.
  • step 430 may include the following steps:
  • Step 431 If the queue is not full, control the queue to perform the enqueue operation for the feature vector corresponding to the current image.
  • step 433 if the queue is full, the queue is controlled to perform a dequeue operation at the head of the queue, and perform a queue operation at the end of the queue for the feature vector corresponding to the current image.
  • a queue of specified length N includes N storage locations, and each storage location can be used to store a feature vector that satisfies the normal structure interval.
  • the specified length of the queue can be flexibly adjusted according to the actual needs of the application scenario. For example, for application scenarios with high accuracy requirements, if the number of images of the object to be traversed is large, a larger specified length is set; for electronic For application scenarios where device storage space requirements are high, a smaller specified length is set, which is not limited in this embodiment.
  • the feature vector a 1 is stored to the first storage location in the queue.
  • the feature vector a 2 is stored to the second storage location in the queue.
  • the N-th feature vector a n satisfy the normal structure section, then the N-th feature vector a n memory locations to store queue, at which point the queue is full.
  • the N + 1 feature vector a n + 1 meets the normal structure interval, adhering to the principle of "first in, first out", then the first feature vector a 1 is removed from the queue head and the second a 2 eigenvectors in the first direction to force a first memory location, and so on, the N-th feature vector a n in the queue head moves to the direction of N-1 storage locations, thereby completing a dequeue operation.
  • the Nth storage location in the queue is empty, then the N + 1th feature vector an + 1 is stored from the end of the queue to the Nth storage location, thereby completing the enqueue operation.
  • a queue-based living body detection method is realized, which can not only effectively filter the false judgment of the prosthesis due to the mutation of the facial feature structure in the prosthesis attack sample as a living body, but also can be applied to various facial features
  • the same crowd that is, the feature vectors in different queues can reflect the normal structure of different facial features, making the living body detection method have good adaptability and versatility.
  • step 350 may include the following steps:
  • Step 351 Calculate the average value of the feature vectors in the feature sequence.
  • Step 353 Calculate the relative change rate of the feature vector corresponding to the current image according to the average value and the feature vector corresponding to the current image.
  • the aspect ratio of the eyes when blinking occurs, the aspect ratio of the eyes will jump obviously, that is, by recording the trend of the aspect ratio of the eyes during the historical image acquisition time period, it is judged whether it has occurred Blink.
  • the determination threshold is set to 0.15, and when the aspect ratio of the eyes is less than 0.15, it is considered that the blinking behavior of the object to be detected is captured once.
  • the decision threshold can be flexibly set according to the actual needs of the application scenarios. For example, for application scenarios that require high detection sensitivity, a larger decision threshold is set, which is not limited in this embodiment .
  • the movement behavior of the living body can be sharply captured by the aspect ratio of the eye.
  • the attacker quickly occludes the outline of the eye in the image multiple times in a row, because several key points of the eye in the image are destroyed , It is easy to cause the aspect ratio of the eye to be less than the judgment threshold, which leads to the false judgment of the prosthesis as a living body.
  • the probability of an apparent jump in the aspect ratio of the eye will be greatly reduced. Extreme cases may also occur when the object to be detected opens the eye.
  • the corresponding eye aspect ratio is already smaller than the determination threshold, which results in the inability to detect the apparent jump of the eye aspect ratio during the blinking of the object to be detected, and the living body is mistakenly judged as a prosthesis.
  • the living body detection method is implemented by the relative change of the facial feature structure of the object to be detected in the image, as shown in the following formula (2):
  • ⁇ represents the relative change rate of the feature vector corresponding to the current image
  • Ear_ave represents the average value of the feature vectors in the feature sequence
  • Ear ' reflects the structure of the facial features of the object to be detected in the current image acquisition time period.
  • the relative change rate ⁇ is not zero, it means that the facial features reflected by Ear 'within the same period of time as the image to be detected is acquired Compared with the normal structure of the facial features reflected by Ear_ave, the structure of the image has changed. Then, the object to be detected may have action behaviors.
  • Step 355 if the relative change rate of the feature vector corresponding to the current image is less than the change threshold, it is captured that the object to be detected has an action behavior.
  • the change threshold is set, that is, only when the relative change rate ⁇ is smaller than the set change threshold, it is considered that there is an action behavior of the object to be detected.
  • the change threshold can be flexibly set according to the actual needs of the application scenario. For example, for an application scenario that requires high detection sensitivity, a larger change threshold is set, which is not limited in this embodiment .
  • the relative change judgment is used instead of the absolute change judgment to avoid facial features with different structures due to different motion behaviors.
  • the blinking amplitude of small eyes is smaller than that of large eyes, and the living body is mistaken for a prosthesis. Defects, which enhances the robustness and stability of live detection.
  • step 355 may include the following steps:
  • Step 3551 If the relative change rate of the feature vector corresponding to the current image is less than the change threshold, control the counter to accumulate.
  • Step 3553 when the count value of the counter exceeds the accumulation threshold, it is captured that the object to be detected has an action behavior.
  • the counter and the accumulation threshold are set only when the accumulated count value in the counter exceeds The accumulation threshold is considered as the action behavior of the object to be detected.
  • step 370 the method described above may further include the following steps:
  • a face recognition model is called to perform face recognition on the image of the object to be detected.
  • the face recognition process may include the following steps:
  • Step 371 Call the face recognition model to extract face features from the image of the object to be detected.
  • the face recognition model is generated based on the model training of the convolutional neural network model.
  • the forward propagation of the face recognition model can be used to obtain the model output, that is, the facial features of the image of the object to be detected can be extracted.
  • Step 373 Perform similarity calculation on the specified facial features according to the extracted facial features.
  • the designated facial features are obtained by extracting facial features from the images of designated persons based on the face recognition model.
  • the designated person may be a person who needs access authorization in a certain building, or a target that a security inspection department needs to track, or a person who needs to be authenticated by a bank.
  • This embodiment does not Types are specifically limited.
  • different types of designated personnel may correspond to different application scenarios, which include but are not limited to: access control authorization scenarios, video surveillance scenarios, face-swapping payment scenarios, etc. Therefore, the face recognition provided in this embodiment
  • the method can be applied to different application scenarios according to different types of designated personnel.
  • the association relationship between the designated face features and the known identity can be established for the designated person for subsequent identification of the designated person Face features are used for identity association.
  • the similarity operation refers to calculating the similarity between the facial features of the image of the object to be detected and the designated facial features.
  • Step 375 Generate a recognition result of the object to be detected according to the similarity calculation result.
  • the operation result is used to indicate the similarity between the facial features of the image of the object to be detected and the specified facial features.
  • the recognition result indicates that the object to be detected has not passed the verification of the identity associated with the specified facial features. At this time, it is regarded as the identity verification of the detected object failure.
  • the recognition result indicates that the object to be detected passes the verification of the identity associated with the specified facial features. At this time, it is regarded as the identity of the object to be detected The verification is successful.
  • the calculation result is used to indicate the similarity between the face features of the image of the object to be detected and the multiple specified face features, that is, the calculation result is A set of multiple similarities, each similarity corresponding to a specified face feature.
  • the designated facial feature with the highest similarity is regarded as the facial feature description of the designated person obtained through identity recognition, and thus the similarity can be obtained according to the association between the designated facial feature and the identity
  • the designated person's identity associated with the largest designated face feature is determined to be the identity of the object to be detected.
  • the process of face recognition is realized, which can be applied to application scenarios that require identity verification / identification.
  • the implementation environment includes a payment user 510, a smartphone 530, and a payment server 550.
  • the payment user 510 brushes the face through the camera configured on the smartphone 530, so that the smartphone 530 obtains the corresponding user image to be recognized by the payment user 510, and then uses the face recognition model to carry out the image of the user to be recognized Face recognition.
  • the user feature of the user image to be recognized is extracted through the face recognition model, and the similarity between the user feature and the specified user feature is calculated. If the similarity is greater than the similarity threshold, the paying user 510 passes the identity verification.
  • the specified user characteristics are extracted by the smartphone 530 for the paying user 510 in advance through the face recognition model.
  • the smartphone 530 After the payment user 510 passes the identity verification, the smartphone 530 initiates an order payment request to the payment server 550 for the order to be paid, thereby completing the payment process of the order to be paid.
  • 15 is a schematic diagram of an implementation environment based on identity recognition in an application scenario. For example, in video surveillance, through identification, the tracking target is determined in multiple face images displayed on the image screen. Many-to-one feature comparison is implemented in this application scenario, which can be regarded as a special case of one-to-one feature comparison.
  • the implementation environment includes a monitoring screen 610, cameras 630 arranged everywhere, and a monitoring server 650 that enables interaction between the camera 630 and the monitoring screen 610.
  • a large number of cameras 630 are arranged to facilitate video surveillance at any time through the image collected by the camera 630.
  • a large number of cameras 630 are arranged to form a video surveillance system, and the image screen is obtained through the interaction between the monitoring server 650 and each camera 630 in the video surveillance system, and then the video surveillance of the tracking target is realized through the image screen on the monitoring screen 610 .
  • the monitoring server 650 For the face recognition of the monitored object in the image frame to determine the tracking target, the monitoring server 650 completes.
  • the face features of multiple face images in the image frame are extracted through the face recognition model, and the similarity between these face features and the specified target feature is calculated separately.
  • the specified target features are extracted in advance based on the tracking target through the face recognition model.
  • the facial features with the highest similarity and the similarity exceeding the similarity threshold can be obtained, and the identity of the monitored object is determined as the identity associated with the facial features with the largest similarity and the similarity exceeding the similarity threshold.
  • the tracking target is identified in the image frame, so as to facilitate continuous tracking of the tracking target.
  • the implementation environment includes a reception device 710, an identification server 730, and an access control device 750.
  • a camera is installed on the reception device 710 to take a face photograph of the entry-exit object 770, and send the obtained image of the person to be identified of the entry-exit object 770 to the recognition server 730 for face recognition.
  • access objects 770 include staff and visitors.
  • the recognition server 730 extracts the person feature of the person to be recognized image through the face recognition model, and calculates the similarity between the person feature and multiple specified person features to obtain the specified person feature with the greatest similarity, and then the specified person feature with the greatest similarity
  • the identity of the associated person is determined as the identity of the person 770 entering and exiting the object, thereby completing the identification of the object 770 entering and exiting.
  • the designated person characteristics are extracted by the recognition server 730 for the entry-exit object 770 in advance through the face recognition model.
  • the recognition server 730 sends an access authorization instruction to the access control device 750 for the access object 770, so that the access control device 750 configures the corresponding access control authority for the access object 770 according to the access authorization instruction, thereby enabling the access object 770 uses this access authority to control the access barriers in the designated work area to perform the release action.
  • the identification server 730 and the access control device 750 can be deployed as the same server, or the reception device 710 and the access control device 750 can be deployed on the same server. This application scenario does not limit this.
  • the living body detection device can be used as a precursor module for identity verification / identification.
  • the living body detection device can accurately determine whether the object to be detected is a living body, and thus realize the filtering of the prosthesis attack sample, which can not only provide sufficient authentication / identity Recognition security, but also can effectively reduce the working pressure and flow pressure of the face recognition model, so as to better facilitate various face recognition tasks.
  • the amount of computer programs involved in the living body detection device is light, and the hardware configuration requirements of the electronic equipment are simple. It can be applied not only to smart phones, but also to servers equipped with the windows operating system and linux operating system, which has fully improved The versatility and practicality of biopsy methods.
  • the following is an embodiment of the device of the present application, which can be used to execute the living body detection method involved in any embodiment of the present application.
  • the device embodiments of the present application please refer to the method embodiments of the living body detection method involved in the present application.
  • a living body detection device 900 includes but is not limited to: an image traversal module 910, a feature extraction module 930, a behavior capturing module 950, and a living body detection module 970.
  • the image traversal module 910 is used to traverse multiple images of the object to be detected, and the currently traversed image is used as the current image.
  • the feature extraction module 930 is used to perform facial feature extraction on the current image to obtain a feature vector corresponding to the current image, and the feature vector is used to describe a structure of a facial feature of the object to be detected in the current image .
  • the behavior capturing module 950 is used to capture the behavior of the object to be detected according to the change of the feature vector corresponding to the current image relative to the feature vector corresponding to the historical image in the feature sequence, the historical image is the For the traversed image, the feature sequence contains at least one feature vector corresponding to the historical image.
  • the living body detection module 970 is configured to determine that the object to be detected is a living body if it is captured that the object to be detected has an action behavior.
  • the feature extraction module includes: an image recognition unit for recognizing facial features in the current image to obtain several key points of facial features in the current image; feature calculation The unit is configured to calculate the structural distance ratio of the facial features according to several key points of the facial features in the current image to obtain the feature vector corresponding to the current image.
  • the facial features in the current image are human face features
  • the image recognition unit includes: a gray-scale processing subunit for performing gray-scale processing on the current image to obtain the current The grayscale image of the image; the model calling subunit is used to input the grayscale image of the current image into a key point model of the human face to perform facial feature recognition, and obtain several key points of facial features in the current image.
  • facial features in the current image include eyes and / or mouth
  • feature vectors corresponding to the current image include eye aspect ratio and / or mouth aspect ratio.
  • the device further includes: a vector comparison module for comparing the feature vector corresponding to the current image with a normal structure interval; a vector addition module for if the feature corresponding to the current image If the vector is within the normal structure interval, the feature vector corresponding to the current image is added to the feature sequence.
  • the feature sequence is a queue of a specified length;
  • the vector addition module includes: a first enqueuing unit for controlling the queue to the current image if the queue is not full The corresponding feature vector performs the enqueue operation; the second enqueue unit is used to control the queue to perform the dequeue operation at the head of the queue if the queue is full, and the feature vector corresponding to the current image at the end of the queue Perform enqueue operations.
  • the behavior capturing module includes: an average value calculation unit for calculating the average value of the feature vectors in the feature sequence; a rate of change calculation unit for calculating the average value and the current The feature vector corresponding to the image, calculating the relative change rate of the feature vector corresponding to the current image; the judging unit is used to capture the object to be detected if the relative change rate of the feature vector corresponding to the current image is less than the change threshold There is action behavior.
  • the judgment unit includes: an accumulation subunit for controlling the counter to accumulate if the relative change rate of the feature vector corresponding to the current image is less than the change threshold; a capture subunit is used for When the count value of the counter exceeds the accumulation threshold, it is captured that there is an action behavior of the object to be detected.
  • the apparatus further includes: a face recognition module, configured to call a face recognition model to perform face recognition on the image of the object to be detected if the object to be detected is a living body.
  • the face recognition module includes: a feature extraction unit for calling the face recognition model to perform facial feature extraction on the image of the object to be detected; a similarity calculation unit for Perform a similarity calculation on the specified facial features according to the extracted facial features; a recognition result generation unit is used to generate a recognition result of the object to be detected according to the calculation results.
  • the living body detection device provided in the above embodiment performs the living body detection process
  • only the above-mentioned division of each functional module is used as an example for illustration.
  • the above-mentioned functions may be allocated by different functional modules as needed.
  • Completed, that is, the internal structure of the living body detection device will be divided into different functional modules to complete all or part of the functions described above.
  • the living body detection device and the living body detection method embodiments provided in the above embodiments belong to the same concept, and the specific manner in which each module performs operations has been described in detail in the method embodiments, and will not be repeated here.
  • an electronic device 1000 includes at least one processor 1001, at least one memory 1002, and at least one communication bus 1003.
  • the memory 1002 stores computer readable instructions, and the processor 1001 reads the computer readable instructions stored in the memory 1002 through the communication bus 1003.
  • a computer-readable storage medium has stored thereon a computer program, and when the computer program is executed by a processor, the living body detection method in each of the foregoing embodiments is implemented.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Computer Interaction (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Geometry (AREA)
  • Ophthalmology & Optometry (AREA)
  • Image Analysis (AREA)
  • Collating Specific Patterns (AREA)

Abstract

本申请实施例公开了一种活体检测方法、装置、电子设备、存储介质及应用活体检测方法的支付系统、视频监控系统、门禁系统，所述活体检测方法包括：对待检测对象的多个图像进行遍历，以当前遍历到的图像作为当前图像；对所述当前图像进行脸部特征提取，得到所述当前图像对应的特征向量，所述特征向量用于描述所述待检测对象在所述当前图像中脸部特征的结构；根据所述当前图像对应的特征向量相对特征序列中历史图像所对应特征向量的变化，捕捉所述待检测对象的动作行为，所述历史图像是所述多个图像中已遍历的图像，所述特征序列包含至少一个历史图像对应的特征向量；如果捕捉到所述待检测对象存在动作行为，则判定所述待检测对象为活体。

Description

活体检测方法、装置、电子设备、存储介质及应用活体检测方法的相关系统
本申请要求于2018年10月25日提交国家知识产权局、申请号为201811250025.1，申请名称为“活体检测方法、装置及应用活体检测方法的相关系统”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请实施例涉及脸部特征识别技术领域，尤其涉及一种活体检测方法、装置、电子设备、存储介质及应用活体检测方法的支付系统、视频监控系统、门禁系统。
背景技术
随着生物特征识别技术的发展，生物特征识别被广泛地应用，例如，刷脸支付、视频监控中的人脸识别，以及门禁授权中的指纹识别、虹膜识别等等。生物特征识别也因此而存在着各种各样的威胁，比如攻击者利用伪造的人脸、指纹、虹膜等等进行生物特征识别。
发明内容
本申请各实施例提供一种活体检测方法、装置、电子设备、存储介质及应用活体检测方法的支付系统、视频监控系统、门禁系统。
本申请实施例提供一种活体检测方法，由电子设备执行，包括：对待检测对象的多个图像进行遍历，以当前遍历到的图像作为当前图像；对所述当前图像进行脸部特征提取，得到所述当前图像对应的特征向量，所述特征向量用于描述所述待检测对象在所述当前图像中脸部特征的结构；根据所述当前图像对应的特征向量相对特征序列中历史图像所对应特征向量的变化，捕捉所述待检测对象的动作行为，所述历史图像是所述多个图像中已遍历的图像，所述特征序列包含至少一个历史图像对应的特征向量；如果捕捉到所述待检测对象存在动作行为，则判定所述 待检测对象为活体。
本申请实施例提供一种活体检测装置，包括：图像遍历模块，用于对待检测对象的多个图像进行遍历，以当前遍历到的图像作为当前图像；特征提取模块，用于对所述当前图像进行脸部特征提取，得到所述当前图像对应的特征向量，所述特征向量用于描述所述待检测对象在所述当前图像中脸部特征的结构；行为捕捉模块，用于根据所述当前图像对应的特征向量相对特征序列中历史图像所对应特征向量的变化，捕捉所述待检测对象的动作行为，所述历史图像是所述多个图像中已遍历的图像，所述特征序列包含至少一个历史图像对应的特征向量；活体检测模块，用于如果捕捉到所述待检测对象存在动作行为，则判定所述待检测对象为活体。
本申请实施例提供一种电子设备，包括处理器及存储器，所述存储器上存储有计算机可读指令，所述计算机可读指令被所述处理器执行时实现如上所述的活体检测方法。
本申请实施例提供一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时实现如上所述的活体检测方法。
本申请实施例提供一种支付系统，所述支付系统包括支付终端和支付服务器，其中，所述支付终端，用于采集支付用户的多个图像；所述支付终端包括活体检测装置，用于根据所述支付用户的多个图像确定对应的特征向量，并根据所确定特征向量之间的相对变化捕捉所述支付用户的动作行为，如果捕捉到所述支付用户的动作行为，则判定所述支付用户为活体；当所述支付用户为活体，所述支付终端对所述支付用户进行身份验证，以在所述支付用户通过身份验证时，向所述支付服务器发起支付请求。
本申请实施例提供一种视频监控系统，所述视频监控系统包括监控屏幕、若干摄像头和监控服务器，其中，若干所述摄像头，用于采集监控对象的多个图像；所述监控服务器包括活体检测装置，用于根据所述监控对象的多个图像确定对应的特征向量，并根据所确定特征向量之间的相对变化捕捉所述监控对象的动作行为，如果捕捉到所述监控对象的 动作行为，则判定所述监控对象为活体；当所述监控对象为活体，所述监控服务器对所述监控对象进行身份识别，以获得追踪目标，并在所述监控屏幕中通过图像画面对所述追踪目标进行视频监控。
本申请实施例提供一种门禁系统，所述门禁系统包括接待设备、识别服务器和门禁控制设备，其中，所述接待设备，用于采集出入对象的多个图像；所述识别服务器包括活体检测装置，用于根据所述出入对象的多个图像确定对应的特征向量，并根据所确定特征向量之间的相对变化捕捉所述出入对象的动作行为，如果捕捉到所述出入对象的动作行为，则判定所述出入对象为活体；当所述出入对象为活体，所述识别服务器对所述出入对象进行身份识别，以使所述门禁控制设备为成功完成身份识别的出入对象配置门禁权限，使得该出入对象根据所配置的门禁权限控制指定工作区域的门禁道闸执行放行动作。
应当理解的是，以上的一般描述和后文的细节描述仅是示例性和解释性的，并不能限制本申请实施例。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分，示出了符合本申请的实施例，并与说明书一起用于解释本申请实施例的原理。
图1是根据一示例性实施例示出的一种电子设备的硬件结构框图；
图2是根据一示例性实施例示出的一种活体检测方法的流程图；
图3是根据一示例性实施例示出的另一种活体检测方法的流程图；
图4是图2对应实施例中步骤330在一个实施例的流程图；
图5是图4对应实施例所涉及的图像中眼睛若干关键点的示意图；
图6是图4对应实施例所涉及的眼睛纵横比的变化趋势的示意图；
图7是图4对应实施例中步骤331在一个实施例的流程图；
图8是图7对应实施例所涉及的人脸关键点模型所构建的索引关系的示意图；
图9是图3对应实施例中步骤430在一个实施例的流程图；
图10为图9对应实施例所涉及的队列为图像对应的特征向量执行 入队操作/出队操作的具体实现示意图；
图11是图2对应实施例中步骤350在一个实施例的流程图；
图12是图11对应实施例中步骤355在一个实施例的流程图；
图13是图2对应实施例中步骤370在一个实施例的流程图；
图14是一应用场景中基于身份验证的实施环境的示意图；
图15是一应用场景中基于身份识别的实施环境的示意图；
图16是一应用场景中基于身份识别的另一个实施环境的示意图；
图17是一应用场景中一种活体检测方法的具体时序图；
图18是图17应用场景所涉及的活体检测方法的具体实现示意图；
图19是根据一示例性实施例示出的一种活体检测装置的框图；
图20是根据一示例性实施例示出的一种电子设备的框图。
通过上述附图，已示出本申请明确的实施例，后文中将有更详细的描述，这些附图和文字描述并不是为了通过任何方式限制本申请实施例构思的范围，而是通过参考特定实施例为本领域技术人员说明本申请实施例的概念。
具体实施方式
这里将详细地对示例性实施例进行说明，其示例表示在附图中。下面的描述涉及附图时，除非另有表示，不同附图中的相同数字表示相同或相似的要素。以下示例性实施例中所描述的实施方式并不代表与本申请实施例相一致的所有实施方式。相反，它们仅是与如所附权利要求书中所详述的、本申请实施例的一些方面相一致的装置和方法的例子。
本申请实施例以生物特征为脸部特征为例进行说明。
活体检测方法，是针对待检测对象的图像进行的活体检测，即检测待检测对象在图像中的脸部特征轮廓是否发生了变化，如果检测到待检测对象在图像中的脸部特征轮廓发生了变化，即判定待检测对象为活体。
例如，待检测对象在图像中的脸部特征为眼睛或者嘴巴，当待检测对象眨眼或者张嘴，将造成图像中的脸部特征轮廓发生变化，由此便可 判定待检测对象为活体。
假体攻击样本，是指攻击者所窃取的待检测对象的图像，是攻击者利用眼睛或者嘴巴的活动特点，对所窃取的待检测对象的图像进行恶意地篡改，例如，用笔涂改了图像中眼睛的轮廓，或者，用笔遮挡了图像中嘴巴的轮廓，使得图像中的脸部特征轮廓发生了变化，而造成假体(伪装的待检测对象，即攻击者)眨眼、张嘴等假象，使得假体被误判为活体。
即使活体检测方法基于待检测对象的一段视频，由于活体检测是以图像帧为单位执行的，攻击者仍然可以依赖于假体攻击样本轻易地破解上述活体检测方法，例如，连续多次快速地遮挡图像中脸部特征轮廓，进而使得假体被误判为活体。
由上可知，现有的活体检测方法仍存在对假体攻击样本的防御性较差的缺陷。
为此，本申请实施例特提出了一种活体检测方法，该种活体检测方法能够有效地提高对假体攻击样本的防御性，具有较高的安全性。
该种活体检测方法由计算机程序实现，与之相对应的，所构建的活体检测装置可存储于架构有冯诺依曼体系的电子设备中，以在该电子设备中执行，进而实现待检测对象的活体检测。例如，电子设备可以是智能手机、平板电脑、笔记本电脑、台式电脑、服务器等等，在此并未加以限定。
请参阅图1，图1是根据本申请一示例性实施例示出的一种电子设备的框图。需要说明的是，该种电子设备只是一个适配于本申请实施例的示例，不能认为是提供了对本申请实施例的使用范围的任何限制。该种电子设备也不能解释为需要依赖于或者必须具有图1中示出的示例性的电子设备100中的一个或者多个组件。
如图1所示，电子设备100包括存储器101、存储控制器103、一个或多个(图1中仅示出一个)处理器105、外设接口107、射频模块109、定位模块111、摄像模块113、音频模块115、触控屏幕117以及 按键模块119。这些组件通过一条或多条通讯总线/信号线121相互通讯。
其中，存储器101可用于存储计算机程序以及模块，如本申请示例性实施例中的活体检测方法及装置所对应的计算机程序及模块，处理器105通过执行存储在存储器101内的计算机程序，从而执行各种功能以及数据处理，以例如完成本申请任一实施例所述的活体检测方法。
存储器101作为资源存储的载体，可以是随机存储器、例如高速随机存储器、非易失性存储器，如一个或多个磁性存储装置、闪存、或者其它固态存储器。存储方式可以是短暂存储或者永久存储。
外设接口107可以包括至少一有线或无线网络接口、至少一串并联转换接口、至少一输入输出接口以及至少一USB接口等，用于将外部各种输入/输出装置耦合至存储器101以及处理器105，以实现与外部各种输入/输出装置的通信。
射频模块109用于收发电磁波，实现电磁波与电信号的相互转换，从而通过通讯网络与其他设备进行通讯。通信网络包括蜂窝式电话网、无线局域网或者城域网，上述通信网络可以使用各种通信标准、协议及技术。
定位模块111用于获取电子设备100的当前所在的地理位置。定位模块111的实例包括但不限于全球卫星定位系统(GPS)、基于无线局域网或者移动通信网的定位技术。
摄像模块113隶属于摄像头，用于拍摄图片或者视频。拍摄的图片或者视频可以存储至存储器101内，还可以通过射频模块109发送至上位机。例如，利用摄像模块113拍摄待检测对象，以形成待检测对象的图像。
音频模块115向用户提供音频接口，其可包括一个或多个麦克风接口、一个或多个扬声器接口以及一个或多个耳机接口。通过音频接口与其它设备进行音频数据的交互。音频数据可以存储至存储器101内，还可以通过射频模块109发送。
触控屏幕117在电子设备100与用户之间提供一个输入输出界面。具体地，用户可通过触控屏幕117进行输入操作，例如点击、触摸、滑 动等手势操作，以使电子设备100对该输入操作进行响应。电子设备100则将文字、图片或者视频任意一种形式或者组合所形成的输出内容通过触控屏幕117向用户显示输出。
按键模块119包括至少一个按键，用以提供用户向电子设备100进行输入的接口，用户可以通过按下不同的按键使电子设备100执行不同的功能。例如，声音调节按键可供用户实现对电子设备100播放的声音音量的调节。
可以理解，图1所示的结构仅为示意，电子设备100还可包括比图1中所示更多或更少的组件，或者具有与图1所示不同的组件。图1中所示的各组件可以采用硬件、软件或者其组合来实现。
请参阅图2，在本申请一示例性实施例中，一种活体检测方法适用于电子设备，电子设备的结构可以如图1所示。
该种活体检测方法可以由电子设备执行，可以包括以下步骤：
步骤310，对待检测对象的多个图像进行遍历，以当前遍历到的图像作为当前图像。
首先，待检测对象的多个图像，可以是指一段视频，即摄像设备对待检测对象进行了一次摄像所生成的；还可以是指多张照片，即摄像设备对待检测对象进行了连续多次拍照所生成的。
也可以理解为，活体检测可以基于待检测对象的一段视频进行，也可以基于待检测对象的多张照片进行，本实施例并未对此加以限定。
其次，多个图像的获取，可以是实时采集到的多个图像，也可以是预先存储的多个图像，即通过读取缓存区域中一历史时间段采集到的多个图像，本实施例也并未对此进行限定。
换而言之，摄像设备实时采集到待检测对象的多个图像之后，可以针对待检测对象的多个图像实时地进行活体检测，也可以先存储待检测对象的图像待以后再处理。例如，在电子设备处理任务较少的时候进行处理，或者按照检测人员的指示进行处理。
其中，摄像设备可以是摄像机、录像机、或者其它具有图像采集功 能的电子设备，例如，智能手机等。
活体检测是以图像帧为单位执行的。因此，在获取到待检测对象的多个图像之后，便可对待检测对象的多个图像进行遍历，以便于根据遍历到的图像进行活体检测。
在此说明的是，对于待检测对象的多个图像而言，当前遍历到的图像视为当前图像，已遍历的图像视为历史图像。
步骤330，对所述当前图像进行脸部特征提取，得到所述当前图像对应的特征向量，所述特征向量用于描述所述待检测对象在所述当前图像中脸部特征的结构。
待检测对象的脸部特征，例如，可以是眼睛、嘴巴、耳朵、虹膜等等。为此，对应于图像的特征向量，用于实现对待检测对象在该图像中脸部特征的结构的准确描述，进而在信息上唯一地表示待检测对象在图像中脸部特征的结构。
可以理解，对于待检测对象不同的脸部特征，图像中脸部特征的结构将有所区别，进而使得图像对应的特征向量也各不相同。例如，待检测对象的脸部特征为眼睛，则图像对应的特征向量为眼睛纵横比，以此描述待检测对象在图像中眼睛的结构；若待检测对象的脸部特征为嘴巴，则图像对应的特征向量为嘴巴纵横比，以此描述待检测对象在图像中嘴巴的结构。
在此，对于图像对应的特征向量，不再进行一一列举，不同的待检测对象，均有其相应的图像，进而存在相应图像所对应的特征向量，以便于准确地描述不同待检测对象在相应图像中脸部特征的结构。
步骤350，根据所述当前图像对应的特征向量相对特征序列中历史图像所对应特征向量的变化，捕捉所述待检测对象的动作行为。
如前所述，当前遍历到的图像视为当前图像，历史图像，则指的是多个图像中已遍历的图像。
由此，特征序列，是通过对待检测对象在多个图像中的脸部特征进行跟踪而生成的，此特征序列包含至少一个历史图像对应的特征向量。
具体地，如图3所示，特征序列的构建过程可以包括以下步骤：
步骤410，将所述当前图像对应的特征向量与正常结构区间进行比较。
步骤430，如果所述当前图像对应的特征向量在所述正常结构区间之内，则将所述当前图像对应的特征向量添加至所述特征序列。
步骤450，如果所述当前图像对应的特征向量在所述正常结构区间之外，则忽略所述当前图像对应的特征向量，继续遍历多个图像中的后一个图像。
应当理解，对于同一待检测对象而言，其脸部特征的结构相对固定，故而，待检测对象在图像中脸部特征的结构也相对固定，视为正常结构。例如，若待检测对象的脸部特征为眼睛，则眼睛睁开时的眼睛轮廓视为正常结构。
正常结构区间，表示待检测对象在图像中脸部特征的正常结构的波动范围。此正常结构区间可根据应用场景的实际需求灵活地设置，例如，在精度要求较高的应用场景中，设置波动范围较窄的正常结构区间，本实施例并未对此加以限定。
由此，仅在图像对应的特征向量在正常结构区间之内，才被允许添加至特征序列，以此防止异常的特则向量存在于特征序列，充分地保障了特征序列的准确性，进而有利于提升活体检测的准确性。
在此说明的是，随着对待检测对象的多个图像的遍历的持续进行，当遍历到多个图像中的后一个图像时，当前图像即转化为历史图像，而遍历到的后一个图像则更新为当前图像。
因此，特征序列，实质上反映了待检测对象在历史图像中脸部特征的正常结构。也可以理解为，特征序列，用于实现对待检测对象在历史图像采集时间段内脸部特征的正常结构的准确描述。
那么，如果当前图像对应的特征向量相对特征序列中历史图像所对应的特征向量发生了变化，则表明待检测对象在当前图像中脸部特征的结构，相对特征序列所描述的待检测对象在历史图像采集时间段内脸部特征的正常结构发生了变化。
仍以待检测对象的脸部特征为眼睛进行说明，正常结构为眼睛睁开 时的眼睛轮廓，那么，发生了变化的结构即为眨眼时的眼睛轮廓。
在本申请实施例中，待检测对象的动作行为，包括但不限于：眨眼行为、张嘴行为、闭嘴行为、招手行为、跺脚行为等等。
由上可知，对于待检测对象的多个图像，当其中一个图像(当前图像)对应的特征向量相对于前若干个图像(历史图像)对应的特征向量发生了相对变化时，表明待检测对象在图像中的脸部特征轮廓发生了变化，例如，待检测对象眨眼了，此时，即视为捕捉到待检测对象的动作行为，进而判定待检测对象为活体。
步骤370，如果捕捉到所述待检测对象存在动作行为，则判定所述待检测对象为活体。
例如，当捕捉到待检测对象的眨眼行为、张嘴行为、闭嘴行为、招手行为、或者跺脚行为等等，便可判定待检测对象为活体。
通过如上所述的过程，实现了基于特征向量相对变化的活体检测方案，即对于待检测对象的多个图像，仅在其中一个图像对应的特征向量相对于前若干个图像对应的特征向量发生了相对变化，待检测对象才会被判定为活体，以此过滤假体攻击样本中脸部特征轮廓的突变而造成的假体误判，从而有效地提高活体检测方法对假体攻击样本的防御性，具有较高的安全性。
请参阅图4，在本申请一示例性实施例中，步骤330可以包括以下步骤：
步骤331，对所述当前图像中的脸部特征进行识别，得到所述当前图像中脸部特征的若干关键点。
如前所述，特征向量用于描述待检测对象在图像中脸部特征的结构，而图像中脸部特征的结构，实质指的是图像中脸部特征的轮廓。
例如，若待检测对象的脸部特征为眼睛，待检测对象在图像中眼睛的结构即是指图像中眼睛的轮廓。
可以理解，图像中脸部特征的轮廓可视为由一系列像素点构成，在计算特征向量时，没有必要使用构成图像中脸部特征轮廓的全部像素点，为此，本实施例中，通过脸部特征识别，得到图像中脸部特征的若干关 键点。
也可以理解为，图像中脸部特征的若干关键点，用于表征构成图像中脸部特征轮廓的若干关键像素点。
仍以待检测对象的脸部特征为眼睛进行说明，如图5所示，图像中眼睛的六个关键点，分别包括：右眼角的一个关键像素点P1、上眼睑的两个关键像素点P2和P3、左眼角的一个关键像素点P4、下眼睑的两个关键像素点P5和P6。
在本申请实施例中，图像中脸部特征的若干关键点，是由不同的坐标(x，y)进行唯一表示的。
在本申请实施例中，根据不同种类的待检测对象，脸部特征识别可以采用不同的方式实现。
在本申请一实施例中，待检测对象为人时，脸部特征识别可以采用人脸关键点模型实现，即调用人脸关键点模型提取得到图像中脸部特征的若干关键点。
步骤333，根据所述当前图像中脸部特征的若干关键点计算脸部特征的结构距离比例，得到所述当前图像对应的特征向量。
如前所述，如果待检测对象的脸部特征为眼睛，则图像对应的特征向量为眼睛纵横比，以此描述待检测对象在图像中眼睛的结构；如果待检测对象的脸部特征为嘴巴，则图像对应的特征向量为嘴巴纵横比，以此描述待检测对象在图像中嘴巴的结构。
其中，眼睛纵横比表示眼睛的结构距离比例，嘴巴纵横比表示嘴巴的结构距离比例。
仍以待检测对象的脸部特征为眼睛进行说明，如图5所示，通过图像中眼睛的六个关键点，即可基本定位出图像中眼睛的轮廓，进而反映出图像中眼睛的结构。具体地，眼睛纵横比的计算公式如(1)所示：
Figure PCTCN2019112196-appb-000001
其中，EAR表示眼睛纵横比，p 1表示右眼角所在关键像素点的坐标，p 2和p 3分别表示上眼睑所在两个关键像素点的坐标，p 4表示左眼角所在 关键像素点的坐标，p 5和p 6分别表示下眼睑所在两个关键像素点的坐标。
||p1-p4||表示左右眼角所在的一对关键像素点之间的坐标差的范数，同理，||p2-p6||表示上下眼睑所在的其中一对关键像素点之间的坐标差的范数，||p3-p5||表示上下眼睑所在的另一对关键像素点之间的坐标差的范数。
计算公式(1)中，分子表示眼睛上下眼睑之间的垂直距离，分母表示眼睛左右眼角之间的水平距离。应当说明的是，由于分子包含了两组垂直距离，而分母仅包含了一组水平距离，为此，对分母进行了加权，即×2。
在此结合计算公式(1)，对通过眼睛纵横比来捕捉待检测对象的眨眼行为的原理进行说明。
如图6所示，当眼睛睁开时，眼睛纵横比大致恒定，仅在0.25范围上下波动，而一旦发生眨眼、闭眼时，由于垂直距离几乎为零，将使得眼睛纵横比也相应地降低为零，再次睁眼时，眼睛纵横比重新上升至0.25范围，由此，即可表明发生了一次眨眼。
基于上述实施例，实现了对图像中脸部特征结构的准确描述，为后续捕捉待检测对象是否存在动作行为提供了依据，进而使得活体检测得以实现。
进一步地，请参阅图7，在本申请一示例性实施例中，所述当前图像中的脸部特征为人脸特征。其中，人脸特征包括但不限于：眉毛、眼睛、鼻子、嘴巴、耳朵等等。
相应地，步骤331可以包括以下步骤：
步骤3311，对所述当前图像进行灰度处理，得到所述当前图像的灰度图。
步骤3313，将所述当前图像的灰度图输入人脸关键点模型进行脸部特征识别，得到所述当前图像中脸部特征的若干关键点。
人脸关键点模型，实质上为图像中的人脸特征构建了索引关系，以便于通过所构建的索引关系能够从图像中提取得到特定人脸特征的若干关键点。
具体地，对于待检测对象的图像，将其输入至人脸关键点模型之后，图像中的人脸特征的若干关键点即进行了索引标记。例如，如图8所示，图像中，左、右眼睛的六个关键点所标记的索引分别为37～42和43～48，嘴巴的二十个关键点所标记的索引为49～68。
同时，相应地存储进行了索引标记的人脸特征的若干关键点在图像中的坐标，以此为图像中的人脸特征构建了索引与坐标之间的索引关系。
那么，基于索引关系，由索引即可得到人脸特征的若干关键点在图像中的坐标。
在本申请一实施例中，人脸关键点模型，是通过海量的图像样本对指定数学模型进行模型训练生成的。其中，图像样本，是指进行了索引标记的图像。
模型训练，实质上是对指定数学模型的参数加以迭代优化，使得由此参数构建的指定算法函数满足收敛条件。
其中，指定数学模型，包括但不限于：逻辑回归、支持向量机、随机森林、神经网络等机器学习模型。
指定算法函数，包括但不限于：最大期望函数、损失函数等等。
举例来说，随机初始化指定数学模型的参数，根据当前一个图像样本计算随机初始化的参数所构建的损失函数的损失值。
如果损失函数的损失值未达到最小，则更新指定数学模型的参数，并根据后一个图像样本计算更新的参数所构建的损失函数的损失值。
如此迭代循环，直至损失函数的损失值达到最小，即视为损失函数收敛，使得指定数学模型收敛为人脸关键点模型，并停止迭代。
否则，迭代更新指定数学模型的参数，并根据其余图像样本迭代计算所更新参数构建的损失函数的损失值，直至损失函数收敛。
值得一提的是，如果在损失函数收敛之前，迭代次数已经达到迭代阈值，也将停止迭代，以此保证模型训练的效率。
由上可知，利用完成模型训练的人脸关键点模型，便可快速实时地得到图像中人脸特征的若干关键点，充分地保障了活体检测的时效性。
此外，基于人脸关键点模型，对于不同面部表情的人脸特征识别， 都有较好的准确性和稳定性，为后续活体检测提供了准确的依据。
请参阅图9，在本申请一示例性实施例中，所述特征序列为指定长度的队列。
相应地，步骤430可以包括以下步骤：
步骤431，如果所述队列未满，则控制所述队列为所述当前图像对应的特征向量执行入队操作。
步骤433，如果所述队列已满，则控制所述队列在队头执行出队操作，并在队尾为所述当前图像对应的特征向量执行入队操作。
如图10所示，指定长度N的队列包括N个存储位置，每一个存储位置可用于存储一个满足正常结构区间的特征向量。
其中，队列的指定长度可以根据应用场景的实际需求灵活地调整，例如，对于精度要求较高的应用场景，如果遍历的待检测对象的图像数量较多，则设置较大的指定长度；对于电子设备存储空间要求较高的应用场景，则设置较小的指定长度，本实施例并未对此加以限定。
假设待检测对象的图像有2n个，每个图像对应的特征向量为a i，1<＝i<＝2n。
当队列为空，如果第一个特征向量a 1满足正常结构区间，则将特征向量a 1存储至队列中的第一个存储位置。
当队列未满，如果第二个特征向量a 2满足正常结构区间，则将特征向量a 2存储至队列中的第二个存储位置。
以此类推，如果第N个特征向量a n满足正常结构区间，则将特征向量a n存储至队列中的第N个存储位置，此时，队列已满。
当队列已满，如果第N+1个特征向量a n+1满足正常结构区间，秉持“先进先出”的原则，则将第一个特征向量a 1从队头移出队列，并将第二个特征向量a 2沿队头方向移动至第一个存储位置，以此类推，第N个特征向量a n沿队头方向移动至第N-1个存储位置，由此完成出队操作。
此时，队列中的第N个存储位置为空，则将第N+1个特征向量a n+1从队尾存储至第N个存储位置，由此完成入队操作。
由上可知，随着待检测对象的图像的持续采集，限于队列中的存储 位置有限，队列中存储的特征向量将随之实时更新，以此实现滑动窗口式的过滤作用，充分地保证队列所描述的待检测对象在历史图像采集时间段内脸部特征的正常结构的准确性。
基于上述实施例，实现了基于队列的活体检测方法，不仅能够有效地过滤由于假体攻击样本中脸部特征结构的突变而造成的假体误判为活体，而且能够适用于人脸特征各不相同的人群，即不同队列中的特征向量可反映出不同人脸特征的正常结构，使得活体检测方法具有良好的适应性和通用性。
请参阅图11，在本申请一示例性实施例中，步骤350可以包括以下步骤：
步骤351，计算所述特征序列中特征向量的平均值。
步骤353，根据所述平均值和所述当前图像对应的特征向量，计算所述当前图像对应的特征向量的相对变化率。
以眼睛纵横比为例，如图6所示，当发生眨眼时，眼睛纵横比会发生明显的跳变，即通过记录眼睛纵横比在历史图像采集时间段内的变化趋势，来判断是否发生了眨眼。
具体地，设定判定阈值为0.15，当眼睛纵横比小于0.15时，则视为捕捉到一次待检测对象的眨眼行为。
在此说明的是，判定阈值可以根据应用场景的实际需要灵活地设定，例如，对检测敏感度要求较高的应用场景，设定较大的判定阈值，本实施例并未对此构成限定。
可以理解，对于活体来说，通过眼睛纵横比可以敏锐地捕捉到活体所存在的动作行为，然而，如果攻击者连续多次快速遮挡图像中眼睛的轮廓，由于图像中眼睛的若干关键点被破坏，很容易造成眼睛纵横比小于判定阈值的情况，而导致假体被误判为活体。
此外，误判还存在另一种情况，即活体被误判为假体。
对于待检测对象的图像而言，如果图像中的眼睛本身就比较小，眼睛纵横比发生明显跳变的概率将大大减小，极端情况可能还会出现在待检测对象睁开眼睛时，图像所对应的眼睛纵横比就已经小于判定阈值， 而导致无法检测到眼睛纵横比在待检测对象眨眼期间所发生的明显跳变，而将活体误判为假体。
为此，本实施例中，活体检测方法通过待检测对象在图像中的脸部特征结构的相对变化实现，如以下公式(2)所示：
Figure PCTCN2019112196-appb-000002
其中，α表示当前图像对应的特征向量的相对变化率，Ear_ave表示特征序列中特征向量的平均值，Ear'表示当前图像对应的特征向量。
也就是说，通过Ear_ave，反映待检测对象在历史图像采集时间段内脸部特征的正常结构。
通过Ear'，反映待检测对象在当前图像采集时间段内脸部特征的结构。
由于历史图像采集时间段和当前图像采集时间段是连续的，故而，如果相对变化率α不为零，则表明在对待检测图像进行图像采集的同一时间段内，Ear'所反映的脸部特征的结构，相对于Ear_ave所反映的脸部特征的正常结构发生了变化，那么，待检测对象可能存在动作行为。
步骤355，如果所述当前图像对应的特征向量的相对变化率小于变化阈值，则捕捉到所述待检测对象存在动作行为。
如前所述，当相对变化率α不为零时，待检测对象可能存在动作行为。为此，本实施例，设定变化阈值，即仅在相对变化率α小于所设定的变化阈值时，视为捕捉到待检测对象存在动作行为。
在此说明的是，变化阈值可以根据应用场景的实际需要灵活地设定，例如，对检测敏感度要求较高的应用场景，设定较大的变化阈值，本实施例并未对此加以限定。
在上述过程中，以相对变化判定替代绝对变化判定，避免结构不同的脸部特征因动作行为幅度不同，例如，小眼睛的眨眼幅度小于大眼睛的眨眼幅度，而造成的活体误判为假体的缺陷，从而增强了活体检测的鲁棒性和稳定性。
进一步地，请参阅图12，在本申请一示例性实施例中，步骤355可 以包括以下步骤：
步骤3551，如果所述当前图像对应的特征向量的相对变化率小于所述变化阈值，则控制计数器累加。
步骤3553，当所述计数器的计数值超过累加阈值时，捕捉到所述待检测对象存在动作行为。
可以理解，对待检测对象进行图像采集时，可能是待检测对象拍照时闭眼了，而并非真正的眨眼，故而，本实施例中，设置计数器及累加阈值，仅当计数器中累加的计数值超过累加阈值，才视为待检测对象存在动作行为。
通过上述实施例的配合，进一步地排除了因图像本身存在的脸部特征结构变化而误检的可能性。
在本申请一示例性实施例中，步骤370之后，如上所述的方法还可以包括以下步骤：
如果所述待检测对象为活体，则调用人脸识别模型对所述待检测对象的图像进行人脸识别。
具体地，如图13所示，人脸识别过程可以包括以下步骤：
步骤371，调用所述人脸识别模型对所述待检测对象的图像进行人脸特征提取。
本实施例中，人脸识别模型是基于卷积神经网络模型的模型训练生成的。
那么，将待检测对象的图像输入至人脸识别模型之后，便能够利用人脸识别模型的前向传播得到模型输出，即提取得到待检测对象的图像的人脸特征。
步骤373，根据提取到的人脸特征进行指定人脸特征的相似度运算。
首先说明的是，指定人脸特征，是基于人脸识别模型对指定人员的图像进行人脸特征提取得到的。此处，指定人员可以是某座大厦内需要进行门禁授权的人，也可以是某个安检部门需要追踪的目标，还可以是某个银行需要进行身份验证的人，本实施例不对指定人员的类型作具体限定。
相应地，不同类型的指定人员可对应于不同的应用场景，该应用场景包括但不限于：门禁授权场景、视频监控场景、刷脸支付场景等等，因此，本实施例所提供的人脸识别方法可根据指定人员的不同类型而适用于不同的应用场景。
进一步地，由于指定人员的身份是已知的，故而，在提取得到指定人脸特征之后，便能够为指定人员建立指定人脸特征与已知身份之间的关联关系，以供后续对指定人脸特征进行身份关联。
其次，相似度运算是指计算待检测对象的图像的人脸特征与指定人脸特征之间的相似度。
相似度越大，则表示待检测对象的图像的人脸特征与指定人脸特征越相似，也可以理解为，待检测对象的身份是指定人脸特征所关联身份的可能性越高，反之，则可能性越小。
应当说明的是，指定人脸特征与身份之间的关联关系，是在指定人脸特征提取过程中，为已知身份的指定人员预先建立的，进而方便于对指定人脸特征进行的身份关联。
步骤375，根据相似度运算结果生成所述待检测对象的识别结果。
在一对一或多对一的特征比对中，例如，身份验证，运算结果用于指示待检测对象的图像的人脸特征与指定人脸特征之间的相似度。
当相似度小于相似阈值时，说明待检测对象的身份与指定人员身份不同，则识别结果指示待检测对象未通过指定人脸特征相关联身份的验证，此时，视为对待检测对象的身份验证失败。
反之，当相似度大于相似阈值时，说明待检测对象的身份与指定人员身份相同，则识别结果指示待检测对象通过指定人脸特征相关联身份的验证，此时，视为对待检测对象的身份验证成功。
进一步地，在一对多的特征比对中，例如，身份识别，运算结果用于指示待检测对象的图像的人脸特征分别与多个指定人脸特征之间的相似度，即运算结果是多个相似度的集合，每一个相似度对应一个指定人脸特征。
那么，相似度最大的指定人脸特征，即视为通过身份识别所得到的 指定人员的人脸特征描述，由此便能够根据指定人脸特征与身份之间的关联关系，获得与该相似度最大的指定人脸特征关联的指定人员身份，进而将该指定人员的身份确定为待检测对象的身份。
基于上述实施例，实现了人脸识别的过程，可以应用于需要身份验证/身份识别的应用场景。
图14是一应用场景中基于身份验证的实施环境的示意图。如图14所示，该应用场景中，实施环境包括支付用户510、智能手机530和支付服务器550。
针对某个待支付订单，支付用户510通过智能手机530所配置的摄像头进行刷脸，使得智能手机530获得支付用户510相应的待识别用户图像，进而利用人脸识别模型对该待识别用户图像进行人脸识别。
具体地，通过人脸识别模型提取待识别用户图像的用户特征，并计算该用户特征与指定用户特征的相似度，若相似度大于相似阈值，则支付用户510通过身份验证。其中，指定用户特征是智能手机530通过人脸识别模型预先为支付用户510提取的。
在支付用户510通过身份验证之后，智能手机530为待支付订单向支付服务器550发起订单支付请求，以此完成待支付订单的支付流程。
图15是一应用场景中基于身份识别的实施环境的示意图。例如，视频监控中，通过身份识别，在图像画面所显示的多个人脸图像中确定追踪目标。该应用场景中实现了多对一的特征比对，可视为一对一特征比对的特例。
如图15所示，该应用场景中，实施环境包括监控屏幕610、布设于各处的摄像头630、以及实现摄像头630和监控屏幕610二者之间交互的监控服务器650。
在该应用场景中，无论是室内还是室外，均布设了大量的摄像头630，以便于随时通过摄像头630采集的图像画面而实现视频监控。具体而言，布设的大量摄像头630形成了视频监控系统，通过监控服务器650与视频监控系统中各摄像头630的交互来获得图像画面，进而在监控屏幕610 中通过图像画面实现对追踪目标的视频监控。
对于图像画面中监控对象的人脸识别，以确定追踪目标，则是由监控服务器650完成的。
具体地，通过人脸识别模型提取图像画面中多个人脸图像的人脸特征，并分别计算这些人脸特征与指定目标特征的相似度。其中，指定目标特征是通过人脸识别模型基于追踪目标预先提取的。
由此，便能够获得相似度最大且相似度超过相似阈值的人脸特征，进而将监控对象的身份确定为该相似度最大且相似度超过相似阈值的人脸特征所关联的身份，以此在图像画面中识别出追踪目标，以便于后续进行追踪目标的连续追踪。
需要说明的是，由于并非每一图像画面中都存在追踪目标，故而对于相似度最大的人脸特征，还需要进行相似度比较，以此确保连续追踪的准确性。
图16是一应用场景中基于身份识别的另一实施环境的示意图。如图16所示，该实施环境包括接待设备710、识别服务器730和门禁控制设备750。
其中，接待设备710上安装有摄像头，以对出入对象770进行人脸拍照，并将获得的出入对象770的待识别人员图像发送至识别服务器730进行人脸识别。本应用场景中，出入对象770包括工作人员和来访人员。
识别服务器730通过人脸识别模型提取待识别人员图像的人员特征，并计算该人员特征与多个指定人员特征的相似度，得到相似度最大的指定人员特征，进而将相似度最大的指定人员特征所关联的人员身份确定为出入对象员770的身份，以此完成出入对象770的身份识别。其中，指定人员特征是识别服务器730通过人脸识别模型预先为出入对象770提取的。
待出入对象770的身份识别完成，识别服务器730为出入对象770向门禁控制设备750发送门禁授权指令，使得门禁控制设备750根据该门禁授权指令为出入对象770配置相应的门禁权限，进而使得出入对象 770凭借该门禁权限控制指定工作区域的门禁道闸执行放行动作。
当然，在不同的应用场景，可以根据实际应用需求进行灵活部署，例如，识别服务器730与门禁控制设备750可部署为同一个服务器，或者，接待设备710与门禁控制设备750部署于同一个服务器，本应用场景并未对此加以限定。
在上述三种应用场景中，活体检测装置可作为身份验证/身份识别的前驱模块。
如图17～图18所示，通过执行步骤801至步骤807，活体检测装置能够准确地判断待检测对象是否为活体，进而实现对假体攻击样本的过滤，不仅能够充分地提供身份验证/身份识别的安全性，而且还能够有效地减轻人脸识别模型的工作压力和流量压力，从而更好地为各种人脸识别任务提供便利。
此外，活体检测装置所涉及的计算机程序量轻巧，对电子设备自身的硬件配置要求简单，不仅可应用于智能手机，而且适用于配置了windows操作系统和linux操作系统的服务器中，充分地提高了活体检测方法的通用性和实用性。
下述为本申请装置实施例，可以用于执行本申请任一实施例所涉及的活体检测方法。对于本申请装置实施例中未披露的细节，请参照本申请所涉及的活体检测方法的方法实施例。
请参阅图19，在本申请一示例性实施例中，一种活体检测装置900包括但不限于：图像遍历模块910、特征提取模块930、行为捕捉模块950和活体检测模块970。
其中，图像遍历模块910用于对待检测对象的多个图像进行遍历，以当前遍历到的图像作为当前图像。
特征提取模块930用于对所述当前图像进行脸部特征提取，得到所述当前图像对应的特征向量，所述特征向量用于描述所述待检测对象在所述当前图像中脸部特征的结构。
行为捕捉模块950用于根据所述当前图像对应的特征向量相对特征序列中历史图像所对应特征向量的变化，捕捉所述待检测对象的动作行为，所述历史图像是所述多个图像中已遍历的图像，所述特征序列包含至少一个历史图像对应的特征向量。
活体检测模块970用于如果捕捉到所述待检测对象存在动作行为，则判定所述待检测对象为活体。
在一示例性实施例中，所述特征提取模块包括：图像识别单元，用于对所述当前图像中的脸部特征进行识别，得到所述当前图像中脸部特征的若干关键点；特征计算单元，用于根据所述当前图像中脸部特征的若干关键点计算脸部特征的结构距离比例，得到所述当前图像对应的特征向量。
在一示例性实施例中，所述当前图像中的脸部特征为人脸特征，所述图像识别单元包括：灰度处理子单元，用于对所述当前图像进行灰度处理，得到所述当前图像的灰度图；模型调用子单元，用于将所述当前图像的灰度图输入人脸关键点模型进行脸部特征识别，得到所述当前图像中脸部特征的若干关键点。
在一示例性实施例中，所述当前图像中的脸部特征包括眼睛和/或嘴巴，所述当前图像对应的特征向量包括眼睛纵横比和/或嘴巴纵横比。
在一示例性实施例中，所述装置还包括：向量比较模块，用于将所述当前图像对应的特征向量与正常结构区间进行比较；向量添加模块，用于如果所述当前图像对应的特征向量在所述正常结构区间之内，则将所述当前图像对应的特征向量添加至所述特征序列。
在一示例性实施例中，所述特征序列为指定长度的队列；所述向量添加模块包括：第一入队单元，用于如果所述队列未满，则控制所述队列为所述当前图像对应的特征向量执行入队操作；第二入队单元，用于如果所述队列已满，则控制所述队列在队头执行出队操作，并在队尾为所述当前图像对应的特征向量执行入队操作。
在一示例性实施例中，所述行为捕捉模块包括：平均值计算单元，用于计算所述特征序列中特征向量的平均值；变化率计算单元，用于根 据所述平均值和所述当前图像对应的特征向量，计算所述当前图像对应的特征向量的相对变化率；判断单元，用于如果所述当前图像对应的特征向量的相对变化率小于变化阈值，则捕捉到所述待检测对象存在动作行为。
在一示例性实施例中，所述判断单元包括：累加子单元，用于如果所述当前图像对应的特征向量的相对变化率小于所述变化阈值，则控制计数器累加；捕捉子单元，用于当所述计数器的计数值超过累加阈值时，捕捉到所述待检测对象存在动作行为。
在一示例性实施例中，所述装置还包括：人脸识别模块，用于如果所述待检测对象为活体，则调用人脸识别模型对所述待检测对象的图像进行人脸识别。
在一示例性实施例中，所述人脸识别模块包括：特征提取单元，用于调用所述人脸识别模型对所述待检测对象的图像进行人脸特征提取；相似度计算单元，用于根据提取到的人脸特征进行指定人脸特征的相似度运算；识别结果生成单元，用于根据运算结果生成所述待检测对象的识别结果。
需要说明的是，上述实施例所提供的活体检测装置在进行活体检测处理时，仅以上述各功能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模块完成，即活体检测装置的内部结构将划分为不同的功能模块，以完成以上描述的全部或者部分功能。
另外，上述实施例所提供的活体检测装置与活体检测方法的实施例属于同一构思，其中各个模块执行操作的具体方式已经在方法实施例中进行了详细描述，此处不再赘述。
请参阅图20，在本申请一示例性实施例中，一种电子设备1000，包括至少一处理器1001、至少一存储器1002、以及至少一通信总线1003。
其中，存储器1002上存储有计算机可读指令，处理器1001通过通信总线1003读取存储器1002中存储的计算机可读指令。
该计算机可读指令被处理器1001执行时实现上述各实施例中的活体检测方法。
在本申请一示例性实施例中，一种计算机可读存储介质，其上存储有计算机程序，该计算机程序被处理器执行时实现上述各实施例中的活体检测方法。
上述内容，仅为本申请的示例性实施例，并非用于限制本申请实施例的实施方案，本领域普通技术人员根据本申请实施例的主要构思和精神，可以十分方便地进行相应的变通或修改，故本申请实施例的保护范围应以权利要求书所要求的保护范围为准。

Claims (17)

  1. 一种活体检测方法，由电子设备执行，包括：
    对待检测对象的多个图像进行遍历，以当前遍历到的图像作为当前图像；
    对所述当前图像进行脸部特征提取，得到所述当前图像对应的特征向量，所述特征向量用于描述所述待检测对象在所述当前图像中脸部特征的结构；
    根据所述当前图像对应的特征向量相对特征序列中历史图像所对应特征向量的变化，捕捉所述待检测对象的动作行为，所述历史图像是所述多个图像中已遍历的图像，所述特征序列包含至少一个历史图像对应的特征向量；
    如果捕捉到所述待检测对象存在动作行为，则判定所述待检测对象为活体。
  2. 如权利要求1所述的方法，所述对所述当前图像进行脸部特征提取，得到所述当前图像对应的特征向量，包括：
    对所述当前图像中的脸部特征进行识别，得到所述当前图像中脸部特征的若干关键点；
    根据所述当前图像中脸部特征的若干关键点计算脸部特征的结构距离比例，得到所述当前图像对应的特征向量。
  3. 如权利要求2所述的方法，所述当前图像中的脸部特征为人脸特征，所述对所述当前图像中的脸部特征进行识别，得到所述当前图像中脸部特征的若干关键点，包括：
    对所述当前图像进行灰度处理，得到所述当前图像的灰度图；
    将所述当前图像的灰度图输入人脸关键点模型进行脸部特征识别，得到所述当前图像中脸部特征的若干关键点。
  4. 如权利要求3所述的方法，所述当前图像中的脸部特征包括眼睛和/或嘴巴，所述当前图像对应的特征向量包括眼睛纵横比和/或嘴巴纵横比。
  5. 如权利要求1所述的方法，所述对所述当前图像进行脸部特征 提取，得到所述当前图像对应的特征向量之后，所述方法还包括：
    将所述当前图像对应的特征向量与正常结构区间进行比较；
    如果所述当前图像对应的特征向量在所述正常结构区间之内，则将所述当前图像对应的特征向量添加至所述特征序列。
  6. 如权利要求5所述的方法，所述特征序列为指定长度的队列；
    所述将所述当前图像对应的特征向量添加至所述特征序列，包括：
    如果所述队列未满，则控制所述队列为所述当前图像对应的特征向量执行入队操作；
    如果所述队列已满，则控制所述队列在队头执行出队操作，并在队尾为所述当前图像对应的特征向量执行入队操作。
  7. 如权利要求1所述的方法，所述根据所述当前图像对应的特征向量相对特征序列中历史图像所对应特征向量的变化，捕捉所述待检测对象的动作行为，包括：
    计算所述特征序列中特征向量的平均值；
    根据所述平均值和所述当前图像对应的特征向量，计算所述当前图像对应的特征向量的相对变化率；
    如果所述当前图像对应的特征向量的相对变化率小于变化阈值，则捕捉到所述待检测对象存在动作行为。
  8. 如权利要求7所述的方法，所述如果所述当前图像对应的特征向量的相对变化率小于变化阈值，则捕捉到所述待检测对象存在动作行为，包括：
    如果所述当前图像对应的特征向量的相对变化率小于所述变化阈值，则控制计数器累加；
    当所述计数器的计数值超过累加阈值时，捕捉到所述待检测对象存在动作行为。
  9. 如权利要求1至8任一项所述的方法，还包括：
    如果所述待检测对象为活体，则调用人脸识别模型对所述待检测对象的图像进行人脸识别。
  10. 如权利要求9所述的方法，所述调用人脸识别模型对所述待检 测对象的图像进行人脸识别，包括：
    调用所述人脸识别模型对所述待检测对象的图像进行人脸特征提取；
    根据提取到的人脸特征进行指定人脸特征的相似度运算；
    根据运算结果生成所述待检测对象的识别结果。
  11. 一种活体检测装置，包括：
    图像遍历模块，用于对待检测对象的多个图像进行遍历，以当前遍历到的图像作为当前图像；
    特征提取模块，用于对所述当前图像进行脸部特征提取，得到所述当前图像对应的特征向量，所述特征向量用于描述所述待检测对象在所述当前图像中脸部特征的结构；
    行为捕捉模块，用于根据所述当前图像对应的特征向量相对特征序列中历史图像所对应特征向量的变化，捕捉所述待检测对象的动作行为，所述历史图像是所述多个图像中已遍历的图像，所述特征序列包含至少一个历史图像对应的特征向量；
    活体检测模块，用于如果捕捉到所述待检测对象存在动作行为，则判定所述待检测对象为活体。
  12. 如权利要求11所述的装置，所述特征提取模块包括：
    图像识别单元，用于对所述当前图像中的脸部特征进行识别，得到所述当前图像中脸部特征的若干关键点；
    特征计算单元，用于根据所述当前图像中脸部特征的若干关键点计算脸部特征的结构距离比例，得到所述当前图像对应的特征向量。
  13. 一种支付系统，包括支付终端和支付服务器，其中，
    所述支付终端，用于采集支付用户的多个图像；
    所述支付终端包括活体检测装置，用于根据所述支付用户的多个图像确定对应的特征向量，并根据所确定特征向量之间的相对变化捕捉所述支付用户的动作行为，如果捕捉到所述支付用户的动作行为，则判定 所述支付用户为活体；
    当所述支付用户为活体，所述支付终端对所述支付用户进行身份验证，以在所述支付用户通过身份验证时，向所述支付服务器发起支付请求。
  14. 一种视频监控系统，包括监控屏幕、若干摄像头和监控服务器，其中，
    若干所述摄像头，用于采集监控对象的多个图像；
    所述监控服务器包括活体检测装置，用于根据所述监控对象的多个图像确定对应的特征向量，并根据所确定特征向量之间的相对变化捕捉所述监控对象的动作行为，如果捕捉到所述监控对象的动作行为，则判定所述监控对象为活体；
    当所述监控对象为活体，所述监控服务器对所述监控对象进行身份识别，以获得追踪目标，并在所述监控屏幕中通过图像画面对所述追踪目标进行视频监控。
  15. 一种门禁系统，包括接待设备、识别服务器和门禁控制设备，其中，
    所述接待设备，用于采集出入对象的多个图像；
    所述识别服务器包括活体检测装置，用于根据所述出入对象的多个图像确定对应的特征向量，并根据所确定特征向量之间的相对变化捕捉所述出入对象的动作行为，如果捕捉到所述出入对象的动作行为，则判定所述出入对象为活体；
    当所述出入对象为活体，所述识别服务器对所述出入对象进行身份识别，以使所述门禁控制设备为成功完成身份识别的出入对象配置门禁权限，使得该出入对象根据所配置的门禁权限控制指定工作区域的门禁道闸执行放行动作。
  16. 一种电子设备，包括处理器以及与所述处理器相连接的存储器， 所述存储器中存储有可由所述处理器执行的计算机可读指令，所述处理器执行所述计算机可读指令以完成权利要求1至10任一项所述的方法。
  17. 一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序可由处理器执行以完成权利要求1至10任一项所述的方法。
PCT/CN2019/112196 2018-10-25 2019-10-21 活体检测方法、装置、电子设备、存储介质及应用活体检测方法的相关系统 WO2020083183A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
EP19875568.8A EP3761222B1 (en) 2018-10-25 2019-10-21 Living body detection method and apparatus, electronic device, storage medium and related system using living body detection method
US17/070,435 US11721087B2 (en) 2018-10-25 2020-10-14 Living body detection method and apparatus, electronic device, storage medium, and related system to which living body detection method is applied
US18/333,363 US20230343071A1 (en) 2018-10-25 2023-06-12 Liveness detection
US18/333,357 US20230343070A1 (en) 2018-10-25 2023-06-12 Liveness detection

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201811250025.1 2018-10-25
CN201811250025.1A CN109492550B (zh) 2018-10-25 2018-10-25 活体检测方法、装置及应用活体检测方法的相关系统

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US17/070,435 Continuation US11721087B2 (en) 2018-10-25 2020-10-14 Living body detection method and apparatus, electronic device, storage medium, and related system to which living body detection method is applied

Publications (1)

Publication Number Publication Date
WO2020083183A1 true WO2020083183A1 (zh) 2020-04-30

Family

ID=65691640

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/112196 WO2020083183A1 (zh) 2018-10-25 2019-10-21 活体检测方法、装置、电子设备、存储介质及应用活体检测方法的相关系统

Country Status (4)

Country Link
US (3) US11721087B2 (zh)
EP (1) EP3761222B1 (zh)
CN (1) CN109492550B (zh)
WO (1) WO2020083183A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111914763A (zh) * 2020-08-04 2020-11-10 网易（杭州）网络有限公司 活体检测方法、装置和终端设备

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109492550B (zh) * 2018-10-25 2023-06-06 腾讯科技（深圳）有限公司 活体检测方法、装置及应用活体检测方法的相关系统
CN110069996A (zh) * 2019-03-22 2019-07-30 北京字节跳动网络技术有限公司 头部动作识别方法、装置和电子设备
WO2020213166A1 (ja) * 2019-04-19 2020-10-22 富士通株式会社 画像処理装置、画像処理方法、及び画像処理プログラム
CN111739060A (zh) * 2019-08-01 2020-10-02 北京京东尚科信息技术有限公司 识别方法、设备及存储介质
EP3905123A3 (en) * 2020-12-11 2022-01-05 Beijing Baidu Netcom Science And Technology Co. Ltd. Human body identification method and apparatus, electronic device, storage medium and program product
CN115622730A (zh) * 2022-08-25 2023-01-17 支付宝(杭州)信息技术有限公司 人脸攻击检测模型的训练方法、人脸攻击检测方法和装置

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105184246A (zh) * 2015-08-28 2015-12-23 北京旷视科技有限公司 活体检测方法和活体检测系统
CN106897658A (zh) * 2015-12-18 2017-06-27 腾讯科技（深圳）有限公司 人脸活体的鉴别方法和装置
CN108549873A (zh) * 2018-04-19 2018-09-18 北京华捷艾米科技有限公司 三维人脸识别方法和三维人脸识别系统
CN109492551A (zh) * 2018-10-25 2019-03-19 腾讯科技（深圳）有限公司 活体检测方法、装置及应用活体检测方法的相关系统

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104361326A (zh) * 2014-11-18 2015-02-18 新开普电子股份有限公司 一种判别活体人脸的方法
CN105893920B (zh) * 2015-01-26 2019-12-27 阿里巴巴集团控股有限公司 一种人脸活体检测方法和装置
CN105095881B (zh) * 2015-08-21 2023-04-07 小米科技有限责任公司 人脸识别方法、装置及终端
CN107808115A (zh) * 2017-09-27 2018-03-16 联想(北京)有限公司 一种活体检测方法、装置及存储介质
US10692243B2 (en) * 2017-12-03 2020-06-23 Facebook, Inc. Optimizations for dynamic object instance detection, segmentation, and structure mapping
CN109492550B (zh) * 2018-10-25 2023-06-06 腾讯科技（深圳）有限公司 活体检测方法、装置及应用活体检测方法的相关系统
US11341351B2 (en) * 2020-01-03 2022-05-24 15 Seconds of Fame, Inc. Methods and apparatus for facial recognition on a user device

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105184246A (zh) * 2015-08-28 2015-12-23 北京旷视科技有限公司 活体检测方法和活体检测系统
CN106897658A (zh) * 2015-12-18 2017-06-27 腾讯科技（深圳）有限公司 人脸活体的鉴别方法和装置
CN108549873A (zh) * 2018-04-19 2018-09-18 北京华捷艾米科技有限公司 三维人脸识别方法和三维人脸识别系统
CN109492551A (zh) * 2018-10-25 2019-03-19 腾讯科技（深圳）有限公司 活体检测方法、装置及应用活体检测方法的相关系统

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111914763A (zh) * 2020-08-04 2020-11-10 网易（杭州）网络有限公司 活体检测方法、装置和终端设备
CN111914763B (zh) * 2020-08-04 2023-11-28 网易（杭州）网络有限公司 活体检测方法、装置和终端设备

Also Published As

Publication number Publication date
EP3761222B1 (en) 2024-02-28
US11721087B2 (en) 2023-08-08
CN109492550A (zh) 2019-03-19
EP3761222A4 (en) 2021-08-11
US20230343070A1 (en) 2023-10-26
US20230343071A1 (en) 2023-10-26
US20210073516A1 (en) 2021-03-11
CN109492550B (zh) 2023-06-06
EP3761222A1 (en) 2021-01-06

Similar Documents

Publication Publication Date Title
WO2020083111A1 (zh) 活体检测方法、装置、电子设备、存储介质及应用活体检测方法的相关系统
WO2020083183A1 (zh) 活体检测方法、装置、电子设备、存储介质及应用活体检测方法的相关系统
US10515199B2 (en) Systems and methods for facial authentication
US11151819B2 (en) Access control method, access control apparatus, system, and storage medium
US10430679B2 (en) Methods and systems for detecting head motion during an authentication transaction
US20170262472A1 (en) Systems and methods for recognition of faces e.g. from mobile-device-generated images of faces
WO2016127437A1 (zh) 活体人脸验证方法及系统、计算机程序产品
US20200042772A1 (en) People-Credentials Comparison Authentication Method, System and Camera
WO2016172923A1 (zh) 视频检测方法、视频检测系统以及计算机程序产品
US10970953B2 (en) Face authentication based smart access control system
CN110705356A (zh) 功能控制方法及相关设备
CN108764153A (zh) 人脸识别方法、装置、系统和存储介质
WO2021082548A1 (zh) 活体检测方法、装置、服务器和人脸识别设备
Fegade et al. Residential security system based on facial recognition
US11600111B2 (en) System and method for face recognition
Xiao et al. Facial recognition in uncontrolled conditions for information security
CN110572618B (zh) 一种非法拍照行为监控方法、装置及系统
RU2815689C1 (ru) Способ, терминал и система для биометрической идентификации
RU2798179C1 (ru) Способ, терминал и система для биометрической идентификации
Srivastava et al. A Machine Learning and IoT-based Anti-spoofing Technique for Liveness Detection and Face Recognition
CN116597545A (zh) 门锁控制方法、装置、电子设备及存储介质
JP2022552389A (ja) 対象者が未承認である可能性が高いことを示すアラートを生成するためのイベントを管理する方法、装置、システム及びプログラム

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: 19875568

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2019875568

Country of ref document: EP

Effective date: 20201002

NENP Non-entry into the national phase

Ref country code: DE