WO2021046773A1 - 人脸防伪检测方法、装置、芯片、电子设备和计算机可读介质 - Google Patents

人脸防伪检测方法、装置、芯片、电子设备和计算机可读介质 Download PDF

Info

Publication number
WO2021046773A1
WO2021046773A1 PCT/CN2019/105449 CN2019105449W WO2021046773A1 WO 2021046773 A1 WO2021046773 A1 WO 2021046773A1 CN 2019105449 W CN2019105449 W CN 2019105449W WO 2021046773 A1 WO2021046773 A1 WO 2021046773A1
Authority
WO
WIPO (PCT)
Prior art keywords
face
infrared light
data
pixel value
light spot
Prior art date
Application number
PCT/CN2019/105449
Other languages
English (en)
French (fr)
Inventor
吕萌
Original Assignee
深圳市汇顶科技股份有限公司
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 深圳市汇顶科技股份有限公司 filed Critical 深圳市汇顶科技股份有限公司
Priority to PCT/CN2019/105449 priority Critical patent/WO2021046773A1/zh
Priority to CN201980001918.9A priority patent/CN110720105A/zh
Publication of WO2021046773A1 publication Critical patent/WO2021046773A1/zh

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
    • 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

Definitions

  • This application relates to the field of information processing technology, and in particular to a method, device, chip, electronic equipment, and computer readable medium for face anti-counterfeiting detection.
  • Face anti-counterfeiting is an important function in face recognition related projects or products, and its function is to identify whether the detected object is the face of a real person. Anti-counterfeiting detection of the face can resist, such as: photos, face masks or three-dimensional printed face models and other false faces, attacks on the face recognition system, and improve the security of the face recognition system.
  • face anti-counterfeiting detection methods are mainly divided into 2D face live detection and 3D face live detection.
  • 3D face live detection is more reliable than 2D face live detection.
  • the 3D face live detection technology provided by the prior art mainly includes the following types:
  • Two or more pre-set spectral emitters emit spectra to the detected face, where each spectral emitter emits a different spectral wavelength; the preset spectral receivers receive more than two spectral signals returned by the detected face Calculate according to the two or more spectral signals to obtain the face anti-counterfeiting detection result.
  • the detected object performs related facial actions according to the pre-set action requirements; the camera takes pictures of the process of performing facial actions on the detected object, and obtains more than two frames of face image data; performs according to the more than two frames of face image data Calculate to obtain the face anti-counterfeiting detection result.
  • face anti-counterfeiting detection requires the user to actively perform actions to complete it, or face anti-counterfeiting detection needs to be completed based on the light data reflected by the face, resulting in a complex detection process and too long time for users.
  • face anti-counterfeiting detection requires the user to actively perform actions to complete it
  • face anti-counterfeiting detection needs to be completed based on the light data reflected by the face
  • the purpose of some embodiments of the present application is to provide a simple and easy-to-operate method, device, chip, electronic device, and computer-readable medium for face anti-counterfeiting detection.
  • the embodiment of the present application provides a face anti-counterfeiting detection method, including: projecting an infrared light spot on a target face, taking a picture of the target face, and obtaining a face photo with an infrared light spot; Image processing is performed on the face photo, the morphological data of the infrared light spot is obtained from the face photograph with infrared light spot; the calculation is performed according to the morphological data of the infrared light spot, and the anti-counterfeiting detection result of the target human face is obtained.
  • the embodiment of the present application also provides a face anti-counterfeiting detection device, including:
  • the first acquisition module is configured to project an infrared light spot on the target face, take a picture of the target human face, and obtain a face photo with the infrared light spot;
  • the second acquisition module is configured to perform image processing on the face photo with infrared light spot acquired by the first acquisition module, and obtain morphological data of the infrared light spot from the face photo with infrared light spot;
  • the detection module is configured to perform calculation according to the morphological data of the infrared light spot acquired by the second acquisition module to acquire the anti-counterfeiting detection result of the target face.
  • the embodiment of the present application also provides a face anti-counterfeiting detection chip.
  • the face anti-counterfeiting detection chip is communicatively connected with a memory, and the memory stores instructions, and the instructions are executed by the face anti-counterfeiting detection chip to enable all
  • the face anti-counterfeiting detection chip can execute the above-mentioned face anti-counterfeiting detection method.
  • An embodiment of the present application also provides an electronic device, including a memory, and the aforementioned face anti-counterfeiting detection chip, and the memory is in communication connection with the face anti-counterfeiting detection chip.
  • the embodiment of the present application also provides a computer-readable storage medium that stores a computer program, and when the computer program is executed by a processor, the above-mentioned face anti-counterfeiting detection method is implemented.
  • the embodiment of this application is now projecting an infrared light spot on a target face and taking a picture to obtain a face photo with an infrared light spot, and calculate according to the morphological data of the infrared light spot in the face photo to obtain the target person
  • the anti-counterfeiting detection result of the face The difference of the morphological data of infrared light spots on different media is used to realize the anti-counterfeiting detection of the human face.
  • the realization method is simple and convenient for users to operate.
  • the embodiment of the patent application uses the difference of the morphological data of the infrared light spots on different media to realize the anti-counterfeiting detection of the human face, the embodiment of the patent application only needs to obtain one face photo to complete the anti-counterfeiting of the human face.
  • the detection solves the problem that the prior art requires the user to take the initiative to take actions to obtain multiple frames of face photos to complete the anti-counterfeiting detection of the face, which causes complicated user operations and a large amount of calculation for face orientation detection.
  • the anti-counterfeiting detection result of the target face can be obtained by calculating the shape data of the infrared light spot, which is simple and easy to implement and can effectively shorten the detection time .
  • the shape data of the infrared light spot is the pixel value distribution data of the infrared light spot;
  • the calculation according to the shape data of the infrared light spot to obtain the anti-counterfeiting detection result of the target face includes: In the pixel value distribution data of, obtain the first pixel value attenuation data and the second pixel value attenuation data; wherein, the first pixel value attenuation data and the second pixel value attenuation data are obtained after the maximum pixel value data attenuation Pixel value, the attenuation degree of the first pixel value attenuation data is different from the attenuation degree of the second pixel value attenuation data; acquiring the first infrared sub-spot corresponding to the first pixel value attenuation data on the infrared light spot The area of, and the area of the second infrared sub-spot corresponding to the second pixel value attenuation data on the infrared spot; between the area of the first infrared
  • obtaining the first pixel value attenuation data and the second pixel value attenuation data from the pixel value distribution data of the infrared light spot includes: obtaining the maximum pixel value data according to the pixel value distribution data of the infrared light spot; The maximum pixel value data and the preset first pixel value attenuation rule are used to obtain the first pixel value attenuation data; and the second pixel value attenuation data is obtained according to the maximum pixel value data and the preset second pixel value attenuation rule.
  • the pixel value of the first pixel value attenuation data is greater than the pixel value of the second pixel value attenuation data.
  • the morphological data of the infrared light spot is a feature vector obtained by dimensionality reduction of the infrared light spot through principal component analysis; the calculation according to the morphological data of the infrared light spot to obtain the anti-counterfeiting detection result of the target human face includes : Use a pre-trained neural network model to calculate the feature vector to obtain the tag value of the feature vector; compare the tag value of the feature vector with a preset tag value threshold, and obtain the target person according to the comparison result The anti-counterfeiting detection result of the face; wherein the neural network model is obtained by training the feature vector of a real human face and a feature vector of a fake human face, and the label value threshold is based on the real human face in the neural network model The label value of the feature vector and the label value of the feature vector of the dummy face are obtained.
  • the training method of the neural network model includes: obtaining a first face photo set with infrared light spots corresponding to a real human face, and a second face photo set with infrared light spots corresponding to a dummy human face; Principal component analysis, respectively obtain the first feature vector set corresponding to the first set of infrared light spots of face photos, and the second set of feature vector corresponding to the second set of infrared light spots of face photos; adopt the The first feature vector set and the second feature vector set train the neural network to obtain a trained neural network model.
  • the image processing is performed on the face photo with infrared light spots, and the infrared light spot is obtained from the face photo with infrared light spots.
  • the morphological data includes: performing image processing on the face photo with a red-line light spot, selecting a target infrared light spot from the two or more infrared light spots, and acquiring morphological data of the target infrared light spot.
  • the embodiment of the application only needs to perform partial processing on the face photo with red line spots, select the target infrared spot from two or more infrared spots, and complete the face anti-counterfeiting detection according to the morphological data of the target infrared spot, without using global data
  • the processing reduces the amount of calculation and shortens the time for face anti-counterfeiting detection.
  • the embodiment of the present application can be adapted to different needs in different scenarios and is selected reasonably The target infrared light spot, thereby further ensuring the accuracy of the face anti-counterfeiting detection results.
  • the face anti-counterfeiting detection method further includes: performing image processing on the face photo with infrared light spots, and obtaining the position coordinate data of the infrared light spots from the face photo with infrared light spots;
  • the position coordinate data of the light spot and the preset calibration data are calculated using a three-dimensional reconstruction algorithm to obtain the three-dimensional information of the target face; judging whether the target face is a face graphic structure according to the three-dimensional information of the target face;
  • the performing image processing on the face photo with infrared light spots, and obtaining the morphological data of the infrared light spots from the face photo with infrared light spots is specifically: if the target human face is a human face graphic structure, Image processing is performed on the face photo with infrared light spots, and morphological data of the infrared light spots are obtained from the face photo with infrared light spots.
  • the embodiment of the present application first judges whether there is a face graphic structure in the face photo, and only further detects the authenticity of the face when there is a face graphic structure, which further improves the efficiency of face anti-counterfeiting detection.
  • FIG. 1 is a flowchart of a method for anti-counterfeiting detection of a face according to the first embodiment of the present application
  • FIG. 2 is a schematic diagram of uniformly projecting multiple infrared light spots onto the entire face of the target face in step 101 of the face anti-counterfeiting detection method shown in FIG. 1;
  • FIG. 3 is a first flowchart of step 103 in the face anti-counterfeiting detection method shown in FIG. 1;
  • step 103 is a schematic diagram of the correspondence between the shape of the infrared light spot and the pixel value distribution data when the method shown in FIG. 3 is used to implement step 103;
  • FIG. 5 is the second flowchart of step 103 in the method for anti-counterfeiting detection of the face shown in FIG. 1;
  • Fig. 6 is a flowchart of the training method of the neural network model in Fig. 5;
  • FIG. 7 is a flowchart of a method for anti-counterfeiting detection of a face according to a second embodiment of the present application.
  • Fig. 8 is a schematic structural diagram of a face anti-counterfeiting detection device according to a third embodiment of the present application.
  • FIG. 9 is a first structural diagram of the detection module 803 shown in FIG. 8;
  • FIG. 10 is a second structural diagram of the detection module 803 shown in FIG. 8;
  • FIG. 11 is a schematic structural diagram of a face anti-counterfeiting detection device according to a fourth embodiment of the present application.
  • FIG. 12 is a schematic structural diagram of an electronic device in a sixth embodiment of the fourth application of the present application.
  • the first embodiment of the present application relates to a face anti-counterfeiting detection method, which can be applied to electronic devices, such as mobile phones, tablet computers, etc., and will not be repeated here.
  • an infrared spot is projected on the target face, the target face is photographed, and the face photo with infrared spot is obtained; the face photo with infrared spot is image processed, from the face photo with infrared spot Obtain the morphological data of the infrared light spot; calculate according to the morphological data of the infrared light spot to obtain the anti-counterfeiting detection result of the target face.
  • the first embodiment of the present application relates to a face anti-counterfeiting detection method, including:
  • Step 101 Project an infrared light spot on the target face, take a picture of the target face, and obtain a face photo with the infrared light spot.
  • step 101 may specifically project an infrared light spot on the target face through a device such as a dot-matrix projector, and take a picture of the target face through a device such as an infrared lens to obtain a face photo with infrared light spots.
  • a device such as a dot-matrix projector
  • this embodiment does not limit the number and positions of infrared light spots projected on the target face in step 101.
  • the number and positions of infrared light spots projected on the target face can be set according to requirements, for example: Set to project one infrared spot to the cheek of the target face, or as shown in Figure 2, set to project multiple infrared spots to the entire face of the target face uniformly, etc. Each case will not be repeated here.
  • Step 102 Perform image processing on the face photo with infrared light spots, and obtain morphological data of the infrared light spots from the face photo with infrared light spots.
  • step 102 can acquire the morphological data of all infrared light spots; or only part of the infrared light spots can be acquired. Specifically, step 102 can Image processing is performed on the face photo with infrared light spots, the target infrared light spot is selected from multiple infrared light spots, and the morphological data of the target infrared light spot is obtained.
  • step 102 when step 102 only obtains the morphological data of part of the infrared light spots, the embodiment of this patent application does not limit the number and positions of infrared light spots to be obtained. Step 102 can obtain different positions and different numbers of infrared light spots according to actual needs. ⁇ morphological data.
  • the target infrared spot is selected from two or more infrared spots, and the face anti-counterfeiting detection can be completed according to the morphological data of the target infrared spot, without the need to use global data for processing, reducing The amount of calculation is reduced, the time for face anti-counterfeiting detection is shortened, and because the target infrared light spot can be selected from more than two infrared light spots, the embodiment of the application can adapt to different needs in different scenarios, and reasonably select the target infrared light spot, thereby Further ensure the accuracy of the face anti-counterfeiting detection results.
  • Step 103 Calculate according to the morphological data of the infrared light spot, and obtain the anti-counterfeiting detection result of the target face.
  • step 103 can be implemented in two ways:
  • step 103 includes:
  • Step 301 Obtain first pixel value attenuation data and second pixel value attenuation data from the pixel value distribution data of the infrared light spot.
  • the first pixel value attenuation data and the second pixel value attenuation data are pixel values obtained after the maximum pixel value data attenuation, and the attenuation degree of the first pixel value attenuation data is different from that of the second pixel value attenuation data degree.
  • step 103 can be completed through the following steps:
  • the maximum pixel value data is obtained.
  • the present embodiment takes the cross section of the infrared light spot as an example to display the pixel value distribution of the infrared light spot.
  • a longitudinal section or a section of other angles can also be used as an example to display the pixel value distribution of the infrared light spot, and each situation will not be repeated here.
  • the maximum pixel value data obtained according to the pixel value distribution data 4 corresponding to the cross section of the infrared light spot 1 as shown in FIG. 4 is defined as I in this embodiment.
  • the first pixel value attenuation data is acquired according to the maximum pixel value data and the preset first pixel value attenuation rule.
  • the first pixel value attenuation rule can be preset as attenuation At this time, the first pixel value attenuation data is
  • the above is only an example, and this embodiment does not limit the first pixel value attenuation rule.
  • the first pixel value attenuation rule can be set according to requirements, which will not be repeated here.
  • the second pixel value attenuation data is acquired.
  • the second pixel value attenuation rule in this embodiment can be preset as attenuation At this time, the second pixel value attenuation data is
  • this embodiment uses the first pixel value attenuation data to have a larger pixel value than the second pixel value.
  • the pixel value of the pixel value attenuation data is described as an example. However, it is understandable that in actual use, the pixel value of the first pixel value attenuation data can also be smaller than the pixel value of the second pixel value attenuation data. , Can detect true and false faces by detecting the degree of convergence of the shape of the infrared light spot to the pixel value of the center point.
  • the principle is to use the different morphological changes of the infrared light spot on different media. The same principle and concept.
  • Step 302 Obtain the area of the first infrared sub-spot corresponding to the first pixel value attenuation data on the infrared light spot, and the area of the second infrared sub-spot corresponding to the second pixel value attenuation data on the infrared light spot.
  • the position of the first pixel value attenuation data on the pixel value distribution data 4 is 5, and the first infrared sub-spot corresponding to the first pixel value attenuation data on the infrared spot 1 is 3.
  • the area of the first infrared sub-spot 3 can be obtained by calculation.
  • the area of the first infrared sub-spot 3 is represented by S3.
  • the position of the second pixel value attenuation data on the pixel value distribution data 4 is 6, and the second infrared sub-spot corresponding to the second pixel value attenuation data on the infrared spot 1 is 2.
  • the area of the second infrared sub-spot 2 can be obtained by calculation.
  • the area of the second infrared sub-spot 2 is represented by S2.
  • Step 303 The ratio value between the area of the first infrared sub-spot and the area of the second infrared sub-spot is compared with a preset real-person face ratio value interval, and the anti-counterfeiting detection result of the target face is obtained according to the comparison result.
  • the real face ratio range is [r1, r2].
  • R ⁇ [r1, r2] the target face is a real face, otherwise, the target face is a fake face.
  • the ratio between the area of the first infrared sub-spot and the area of the second infrared sub-spot obtained from a real human face through the steps shown in FIG. 3 is R1.
  • the ratio between the area of the first infrared sub-spot and the area of the second infrared sub-spot is R2.
  • the inventor has learned through repeated experiments that for materials with weaker infrared scattering ability than human skin, such as opaque plastic, plaster or paper, R1>R2; for materials with weaker infrared scattering ability than human skin, such as translucent plastics, R1 ⁇ R2.
  • the real-person face ratio value interval can be obtained by using the steps shown in FIG. 3 for different real-person faces.
  • the above is only an example.
  • the scale value interval of a real person's face can also be obtained by other methods, and I will not repeat them here.
  • the three image value data are: first pixel value attenuation data, second pixel value attenuation data, and third pixel value attenuation data, and the first pixel value
  • the pixel value of the attenuation data is greater than the pixel value of the second pixel value attenuation data
  • the pixel value of the second pixel value attenuation data is greater than the pixel value of the third image value attenuation data
  • they are obtained according to the steps shown in Figure 3:
  • the ratio of the third infrared sub-spot corresponding to the third pixel value attenuation data is R
  • this embodiment calculates R21, R32, and R31, and the method of detecting the authenticity of the face based on the calculation result is only an example.
  • the authenticity of the face can also be detected by the probability method, such as :
  • R21, R32 and R31 have two values ⁇ [r1, r2], the target face is a real face, otherwise, the target face is a fake face. Each method will not be repeated here.
  • step 103 when the morphological data of the infrared light spot is a feature vector obtained by reducing the dimensionality of the infrared light spot through principal component analysis, as shown in FIG. 5, step 103 includes:
  • step 501 a pre-trained neural network model is used to calculate the feature vector to obtain the label value of the feature vector.
  • Step 502 Compare the tag value of the feature vector with a preset tag value threshold, and obtain an anti-counterfeiting detection result of the target face according to the comparison result.
  • the neural network model is obtained by training the feature vector of the real face and the feature vector of the fake face; the label value threshold is based on the label value of the feature vector of the real face and the feature of the fake face in the neural network model The label value of the vector is obtained.
  • the training method of the neural network may include:
  • Step 601 Obtain a first face photo set with infrared light spots corresponding to a real human face, and a second face photo set with infrared light spots corresponding to a fake human face.
  • Step 602 Using principal component analysis, a first feature vector set corresponding to the first face photo set with infrared light spots and a second feature vector set corresponding to the second face photo set with infrared light spots are respectively obtained.
  • each infrared light spot in each face photo with infrared light spot in the first pair of face photos with infrared light spot is processed to obtain the corresponding feature vector, and the first pair of facial photos is generated according to the feature vector.
  • the method for obtaining the second feature vector set is the same as the method for obtaining the first feature vector set, and will not be repeated here.
  • Step 603 Use the first feature vector set and the second feature vector set to train the neural network to obtain a trained neural network model.
  • the number of two or more real human faces and the number of two or more fake human faces is not limited. In the actual use process, in order to ensure the nerve as much as possible The reliability of the network model can obtain as many real and fake faces as possible.
  • the label value threshold may be 0.5.
  • the tag value threshold value can be set larger, such as 0.7, 0.75, etc.
  • the tag value threshold can also be set smaller, such as 0.4, 0.35, etc., which will not be repeated here.
  • the embodiment of this application is now projecting an infrared light spot on a target face and taking a picture to obtain a face photo with an infrared light spot, and calculate according to the morphological data of the infrared light spot in the face photo to obtain the target person
  • the anti-counterfeiting detection result of the face uses the difference of the morphological data of the infrared light spot on different media to realize the anti-counterfeiting detection of the human face, the realization method is simple, and the user operation is convenient.
  • the embodiment of the patent application uses the difference of the morphological data of the infrared light spots on different media to realize the anti-counterfeiting detection of the human face, the embodiment of the patent application only needs to obtain one face photo to complete the anti-counterfeiting of the human face.
  • the detection solves the problem that the prior art requires the user to take the initiative to take actions to obtain multiple frames of face photos to complete the anti-counterfeiting detection of the face, which causes complicated user operations and a large amount of calculation for face orientation detection.
  • the second embodiment of the present application relates to a face anti-counterfeiting detection method, which is basically the same as that shown in FIG. 1, with the difference that it also includes:
  • Step 104 Perform image processing on the face photo with infrared light spots, and obtain position coordinate data of the infrared light spots from the face photo with infrared light spots.
  • Step 105 According to the position coordinate data of the infrared light spot and the preset calibration data, a three-dimensional reconstruction algorithm is used for calculation to obtain the three-dimensional information of the target face.
  • Step 106 Judging whether the target human face has a human face graphic structure according to the three-dimensional information of the target human face.
  • step 102 is specifically step 102'. If the target face is a face graphic structure, image processing is performed on the face photo with infrared light spots, and the shape data of the infrared light spot is obtained from the face photos with infrared light spots.
  • the embodiment of this application first judges whether there is a face graphic structure in the face photo, and only when there is a face graphic structure, can the authenticity of the face be further determined. Performing detection further improves the efficiency of face anti-counterfeiting detection.
  • the third embodiment of the present application relates to a face anti-counterfeiting detection device, including:
  • the first acquisition module 801 is configured to project an infrared light spot on a target face, take a picture of the target human face, and obtain a face photo with an infrared light spot;
  • the second acquisition module 802 is configured to perform image processing on the face photo with infrared light spot acquired by the first acquisition module 801, and obtain morphological data of the infrared light spot from the face photo with infrared light spot;
  • the detection module 803 is configured to perform calculation according to the morphological data of the infrared light spot acquired by the second acquisition module 802 to acquire the anti-counterfeiting detection result of the target face.
  • the morphological data of the infrared light spot is pixel value distribution data of the infrared light spot.
  • the detection module 803 may include:
  • the first obtaining sub-module 901 is configured to obtain first pixel value attenuation data and second pixel value attenuation data from the pixel value distribution data of the infrared light spot;
  • the second acquisition sub-module 902 acquires the area of the first infrared sub-spot corresponding to the first pixel value attenuation data acquired by the first acquisition sub-module on the infrared light spot, and the first acquisition sub-module acquired the first infrared sub-spot area The area of the second infrared sub-spot corresponding to the two-pixel value attenuation data on the infrared light spot.
  • the first detection sub-module 903 is configured to compare the ratio between the area of the first infrared sub-spot and the area of the second infrared sub-spot acquired by the second acquisition sub-module 902 with a preset real person The face ratio value interval is compared, and the anti-counterfeiting detection result of the target face is obtained according to the comparison result.
  • the first obtaining sub-module 901 is specifically configured to obtain maximum pixel value data according to the pixel value distribution data of the infrared light spot; according to the maximum pixel value data and a preset first pixel value attenuation rule, obtain First pixel value attenuation data; acquiring the second pixel value attenuation data according to the maximum pixel value data and a preset second pixel value attenuation rule.
  • the morphological data of the infrared light spot is a feature vector obtained by dimensionality reduction of the infrared light spot through principal component analysis.
  • the detection module 803 may include:
  • the fourth acquisition sub-module 1001 is configured to calculate the feature vector using a pre-trained neural network model to obtain the label value of the feature vector;
  • the second detection submodule 1002 is configured to compare the label value of the feature vector acquired by the fourth acquisition submodule 1001 with a preset label value threshold, and acquire the anti-counterfeiting detection result of the target face according to the comparison result;
  • the neural network model is obtained by training the feature vector of a real human face and a feature vector of a fake human face;
  • the label value threshold is obtained according to the label value of the feature vector of the real human face and the label value of the feature vector of the fake human face in the neural network model.
  • the detection module 803 may further include: a neural network model training sub-module 1003;
  • the neural network training sub-module 1003 may include:
  • the first obtaining unit 10031 is configured to obtain a first set of human faces with infrared light spots corresponding to real human faces, and a second set of human faces with infrared light spots corresponding to fake human faces;
  • the second acquiring unit 10032 is configured to use the principal component analysis to separately acquire a first feature vector set corresponding to the first face photo set with infrared light spots and the second face photo set with infrared light spots The corresponding second feature vector set;
  • the training unit 10033 is configured to use the first feature vector set and the second feature vector set to train a neural network, and obtain a trained neural network model.
  • the second acquisition module 802 is also used to perform image processing on the face photo with red light spots, from the two or more infrared light spots. Select the target infrared spot from the infrared spot; obtain the morphological data of the target infrared spot.
  • the embodiment of this application is now projecting an infrared light spot on a target face and taking a picture to obtain a face photo with an infrared light spot, and calculate according to the morphological data of the infrared light spot in the face photo to obtain the target person
  • the anti-counterfeiting detection result of the face uses the difference of the morphological data of the infrared light spot on different media to realize the anti-counterfeiting detection of the human face, the realization method is simple, and the user operation is convenient.
  • the embodiment of the patent application uses the difference of the morphological data of the infrared light spots on different media to realize the anti-counterfeiting detection of the human face, the embodiment of the patent application only needs to obtain one face photo to complete the anti-counterfeiting of the human face.
  • the detection solves the problem that the prior art requires the user to take active actions to obtain multiple frames of face photos to complete the anti-counterfeiting detection of the face, which causes complicated user operations and large amount of calculation for face orientation detection.
  • the fourth embodiment of the present application relates to a face anti-counterfeiting detection device, which is basically the same as that shown in FIG. 8, with the difference that it further includes:
  • the third acquisition module 804 is configured to perform image processing on the face photo with infrared light spots acquired by the first acquisition module 801, and obtain position coordinate data of the infrared light spots from the face photo with infrared light spots;
  • the fourth acquisition module 805 is configured to calculate the three-dimensional reconstruction algorithm according to the position coordinate data of the infrared spot acquired by the third acquisition module 804 and the preset calibration data to acquire the three-dimensional information of the target face;
  • the judging module 806 is configured to judge whether the target face is a face graphic structure according to the three-dimensional information of the target face acquired by the fourth acquiring module 805;
  • the second acquiring module 802 is specifically configured to image the face photo with infrared light spots acquired by the first acquiring module 801 if the determining module 806 determines that the target face is a face graphic structure Processing, obtaining the morphological data of the infrared light spot from the face photo with the infrared light spot.
  • the embodiment of this application first judges whether there is a face graphic structure in the face photo, and only when there is a face graphic structure, can the authenticity of the face be further determined. Performing detection further improves the efficiency of face anti-counterfeiting detection.
  • the fifth embodiment of the present application relates to a face anti-counterfeiting detection chip.
  • the face anti-counterfeiting detection chip is communicatively connected with a memory.
  • the memory stores instructions.
  • the instructions are executed by the face anti-counterfeiting detection chip to enable all
  • the face anti-counterfeiting detection chip can execute the face anti-counterfeiting detection method described in the first embodiment and the second embodiment of the present application.
  • the embodiment of this application is now projecting an infrared light spot on a target face and taking a picture to obtain a face photo with an infrared light spot, and calculate according to the morphological data of the infrared light spot in the face photo to obtain the target person
  • the anti-counterfeiting detection result of the face uses the difference of the morphological data of the infrared light spot on different media to realize the anti-counterfeiting detection of the human face, the realization method is simple, and the user operation is convenient.
  • the embodiment of the patent application uses the difference of the morphological data of the infrared light spot on different media to realize the anti-counterfeiting detection of the human face, the embodiment of the patent application only needs to obtain one face photo to complete the anti-counterfeiting of the human face.
  • the detection solves the problem that the prior art requires the user to take active actions to obtain multiple frames of face photos to complete the anti-counterfeiting detection of the face, which causes complicated user operations and a large amount of calculation for face orientation detection.
  • the sixth embodiment of the present application relates to an electronic device, including: a memory 1201, and the face anti-counterfeiting detection chip 1202 described in the fifth embodiment above, the memory 1201 and the face anti-counterfeiting detection chip 1202
  • the chip 1202 is in communication connection.
  • the embodiment of this application is now projecting an infrared light spot on a target face and taking a picture to obtain a face photo with an infrared light spot, and calculate according to the morphological data of the infrared light spot in the face photo to obtain the target person
  • the anti-counterfeiting detection result of the face uses the difference of the morphological data of the infrared light spot on different media to realize the anti-counterfeiting detection of the human face, the realization method is simple, and the user operation is convenient.
  • the embodiment of the patent application uses the difference of the morphological data of the infrared light spots on different media to realize the anti-counterfeiting detection of the human face, the embodiment of the patent application only needs to obtain one face photo to complete the anti-counterfeiting of the human face.
  • the detection solves the problem that the prior art requires the user to take active actions to obtain multiple frames of face photos to complete the anti-counterfeiting detection of the face, which causes complicated user operations and large amount of calculation for face orientation detection.
  • the seventh embodiment of the present application relates to a computer-readable storage medium that stores a computer program that, when executed by a processor, implements the face anti-counterfeiting detection methods described in the first and second embodiments of the present application. .
  • the embodiment of this application is now projecting an infrared light spot on a target face and taking a picture to obtain a face photo with an infrared light spot, and calculate according to the morphological data of the infrared light spot in the face photo to obtain the target person
  • the anti-counterfeiting detection result of the face uses the difference of the morphological data of the infrared light spot on different media to realize the anti-counterfeiting detection of the human face, the realization method is simple, and the user operation is convenient.
  • the embodiment of the patent application uses the difference of the morphological data of the infrared light spots on different media to realize the anti-counterfeiting detection of the human face, the embodiment of the patent application only needs to obtain one face photo to complete the anti-counterfeiting of the human face.
  • the detection solves the problem that the prior art requires the user to take active actions to obtain multiple frames of face photos to complete the anti-counterfeiting detection of the face, which causes complicated user operations and large amount of calculation for face orientation detection.

Abstract

一种人脸防伪检测方法和装置，涉及信息处理技术。人脸防伪检测方法，包括：向目标人脸投影红外线光斑，对所述目标人脸进行拍照，获取带红外线光斑的人脸照片(101)；对带红外线光斑的人脸照片进行图像处理，从带红外线光斑的人脸照片中获取红外线光斑的形态数据(102)；根据红外线光斑的形态数据进行计算，获取目标人脸的防伪检测结果(103)。上述方法实现简单、易于操作。

Description

人脸防伪检测方法、装置、芯片、电子设备和计算机可读介质 技术领域
本申请涉及信息处理技术领域，特别涉及一种人脸防伪检测方法、装置、芯片、电子设备和计算机可读介质。
背景技术
人脸防伪是人脸识别相关工程或产品里的一项重要功能，其作用是为了识别出被检测对象是否为真实人的人脸。通过人脸防伪检测可以抵抗，如：照片、人脸面具或者三维打印的人脸模型等虚假人脸，对人脸识别系统的攻击，提高人脸识别系统的安全性。
目前，人脸防伪检测方法主要分为2D人脸活体检测和3D人脸活体检测，其中，3D人脸活体检测相比于2D人脸活体检测更为可靠。现有技术提供的3D人脸活体检测技术，主要包括以下几种：
1、多光谱人脸活体检测法
通过预先设置的两个以上光谱发射器向被检测人脸发射光谱，其中，每个光谱发射器发射的光谱波长不同；通过预先设置的光谱接收器接收被检测人脸返回的两个以上光谱信号；根据所述两个以上光谱信号进行计算，获取人脸防伪检测结果。
2、动作配合人脸活体检测法
被检测对象按照预先设置的动作要求执行相关的脸部动作；照相机对被检测对象执行脸部动作的过程进行拍照，获取两帧以上人脸图像数据；根据所述两帧以上人脸图像数据进行计算，获取人脸防伪检测结果。
3、红外线人脸活体检测法
通过预先设置的红外线发射器向被检测人脸发射红外线；通过预先设置的红外线接收器接收被检测人脸反射的红外线；根据所述发射的红外线和所述被检测人脸反射的红外线进行计算，获取人脸防伪检测结果。
发明人发现现有技术至少存在以下问题：人脸防伪检测需要用户主动做动作配合完成，或者人脸防伪检测需要根据人脸反射回的光线数据完成，造成整个检测过程复杂且时间过长，用户使用体验差的问题。
发明内容
本申请部分实施例的目的在于提供一种实现简单、便于操作的人脸防伪检测方法、装置、芯片、电子设备和计算机可读介质。
本申请实施例提供了一种人脸防伪检测方法，包括：向目标人脸投影红外线光斑，对所述目标人脸进行拍照，获取带红外线光斑的人脸照片；对所述带红外线光斑的人脸照片进行图像处理，从所述带红外线光斑的人脸照片中获取红外线光斑的形态数据；根据所述红外线光斑的形态数据进行计算，获取所述目标人脸的防伪检测结果。
本申请实施例还提供了一种人脸防伪检测装置，包括：
第一获取模块，用于向目标人脸投影红外线光斑，对所述目标人脸进行拍照，获取带红外线光斑的人脸照片；
第二获取模块，用于对所述第一获取模块获取的所述带红外线光斑的人脸照片进行图像处理，从所述带红外线光斑的人脸照片中获取红外线光斑的形态数据；
检测模块，用于根据所述第二获取模块获取的所述红外线光斑的形态数据进行计算，获取所述目标人脸的防伪检测结果。
本申请实施例还提供了一种人脸防伪检测芯片，所述人脸防伪检测芯片与存储器通信连接，所述存储器存储有指令，所述指令被所述人脸防伪检测芯片执行，以使所述人脸防伪检测芯片能够执行以上所述的人脸防伪检测方法。
本申请实施例还提供了一种电子设备，包括：存储器，以及以上所述的人脸防伪检测芯片，所述存储器与所述人脸防伪检测芯片通信连接。
本申请实施例还提供了一种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上所述的人脸防伪检测方法。
本申请实施例现对于现有技术而言，在向目标人脸投影红外线光斑并进行拍照，获取带红外线光斑的人脸照片，并根据人脸照片中红外线光斑的形态数据进行计算，获取目标人脸的防伪检测结果。利用了红外线光斑在不同介质上形态数据的差异性实现对人脸的防伪检测，其实现方法简单，并且便于用户操作。由于本专利申请是根据人脸照片中红外线光斑的形态数据进行计算，获取目标人脸的防伪检测结果，而人脸照片是在向目标人脸投影红外线光斑时进行拍照获得的，大大缩短了整个人脸防伪检测过程的时间，解决了现有技术需要根据人脸反射回的光线数据进行人脸防伪检测，造成检测时间过长的问题，也解决了现有技术需要用户主动做动作配合完成人脸防伪检测，动作过程缓慢，造成检测时间过长的问题，由于整个检测时间缩短了，进而提高了用户的使用 体验。另外，由于本专利申请实施例利用了红外线光斑在不同介质上形态数据的差异性实现对人脸的防伪检测，使得本专利申请实施例只需要获取一帧人脸照片即可完成人脸的防伪检测，解决了现有技术需要用户主动做动作，获取多帧人脸照片完成人脸的防伪检测，造成用户操作复杂，并且人脸方位检测计算量大的问题。而且，由于无需根据发射的红外线和被检测人脸反射的红外线进行计算，通过对红外线光斑的形态数据进行计算，即可获取到目标人脸的防伪检测结果，简单易行而且可以有效缩短检测时长。
例如，所述红外线光斑的形态数据为所述红外线光斑的像素值分布数据；所述根据所述红外线光斑的形态数据进行计算，获取所述目标人脸的防伪检测结果包括：从所述红外线光斑的像素值分布数据中，获取第一像素值衰减数据和第二像素值衰减数据；其中，所述第一像素值衰减数据和所述第二像素值衰减数据为最大像素值数据衰减后得到的像素值，所述第一像素值衰减数据的衰减程度不同于所述第二像素值衰减数据的衰减程度；获取所述第一像素值衰减数据在所述红外线光斑上对应的第一红外线子光斑的面积，以及所述第二像素值衰减数据在所述红外线光斑上对应的第二红外线子光斑的面积；将所述第一红外线子光斑的面积与所述第二红外线子光斑的面积之间的比例值与预先设置的真人人脸比例值区间进行比较，根据比较结果获取所述目标人脸的防伪检测结果。
例如：从所述红外线光斑的像素值分布数据中，获取第一像素值衰减数据和第二像素值衰减数据包括：根据所述红外线光斑的像素值分布数据，获取最大像素值数据；根据所述最大像素值数据以及预先设置的第一像素值衰减规则，获取第一像素值衰减数据；根据所述最大像素值数据以及预先设置的第二 像素值衰减规则，获取第二像素值衰减数据。
例如，所述第一像素值衰减数据的像素值大于所述第二像素值衰减数据的像素值。
例如，所述红外线光斑的形态数据为通过主成分分析对所述红外光斑降维获得的特征向量；所述根据所述红外线光斑的形态数据进行计算，获取所述目标人脸的防伪检测结果包括：采用预先训练好的神经网络模型对所述特征向量进行计算，获得特征向量的标签值；将所述特征向量的标签值与预先设置的标签值阈值进行比较，根据比较结果获取所述目标人脸的防伪检测结果；其中，所述神经网络模型是通过真人人脸的特征向量和假人人脸的特征向量训练获得的，所述标签值阈值是根据所述神经网络模型中真人人脸的特征向量的标签值和假人人脸的特征向量的标签值获得的。
例如，所述神经网络模型的训练方法包括：获取真人人脸对应的第一带红外线光斑的人脸照片集，以及假人人脸对应的第二带红外线光斑的人脸照片集；采用所述主成分分析，分别获取所述第一带红外线光斑的人脸照片集对应的第一特征向量集，以及所述第二带红外线光斑的人脸照片集对应的第二特征向量集；采用所述第一特征向量集和所述第二特征向量集对神经网络进行训练，获取训练好的神经网络模型。
例如，如果所述带红外线光斑的人脸照片包括两个以上红外线光斑，所述对所述带红外线光斑的人脸照片进行图像处理，从所述带红外线光斑的人脸照片中获取红外线光斑的形态数据，包括：对所述带红线光斑的人脸照片进行图像处理，从所述两个以上红外线光斑中选取目标红外线光斑；获取所述目标红外线光斑的形态数据。
本申请实施例只需要对带红线光斑的人脸照片进行局部处理，从两个以上红外线光斑中选取目标红外线光斑，并根据目标红外线光斑的形态数据就可以完成人脸防伪检测，无需采用全局数据进行处理，减小了计算量，缩短了人脸防伪检测时间，并且，由于目标红外线光斑可以从两个以上红外线光斑中选取获得，使得本申请实施例可以适应不同场景下的不同需求，合理选取目标红外线光斑，从而进一步保证人脸防伪检测结果的准确性。
例如，所述人脸防伪检测方法，还包括：对所述带红外线光斑的人脸照片进行图像处理，从所述带红外线光斑的人脸照片中获取红外线光斑的位置坐标数据；根据所述红外线光斑的位置坐标数据和预先设置的标定数据，采用三维重建算法进行计算，获取所述目标人脸的三维信息；根据所述目标人脸的三维信息判断所述目标人脸是否为人脸图形结构；所述对所述带红外线光斑的人脸照片进行图像处理，从所述带红外线光斑的人脸照片中获取红外线光斑的形态数据具体为：如果所述目标人脸是人脸图形结构，对所述带红外线光斑的人脸照片进行图像处理，从所述带红外线光斑的人脸照片中获取红外线光斑的形态数据。
本申请实施例先对人脸照片中是否存在人脸图形结构进行了判断，只有当存在人脸图形结构时才进一步对人脸真伪进行检测，进一步提升了人脸防伪检测的效率。
附图说明
一个或多个实施例通过与之对应的附图中的图片进行示例性说明，这些示例性说明并不构成对实施例的限定。
图1是根据本申请第一实施例中的人脸防伪检测方法的流程图；
图2是图1所示的人脸防伪检测方法中步骤101向目标人脸的全脸均匀投影多个红外线光斑的示意图；
图3是图1所示的人脸防伪检测方法中步骤103的流程图一；
图4是采用图3所示的方法实现步骤103时，红外线光斑形状与像素值分布数据的对应关系示意图；
图5是图1所示的人脸防伪检测方法中步骤103的流程图二；
图6是图5中神经网络模型的训练方法的流程图；
图7是根据本申请第二实施例中的人脸防伪检测方法的流程图；
图8是根据本申请第三实施例中的人脸防伪检测装置的结构示意图；
图9是图8所示的检测模块803的结构示意图一；
图10是图8所示的检测模块803的结构示意图二；
图11是根据本申请第四实施例中的人脸防伪检测装置的结构示意图；
图12是本申请第四本申请第六实施例中的电子设备的结构示意图。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请部分实施例进行进一步详细说明。应当理解，此处所描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。
本申请第一实施例涉及一种人脸防伪检测方法，可应用于电子设备，如手机，平板电脑等，在此不一一赘述。在本实施例中，向目标人脸投影红外线光斑，对目标人脸进行拍照，获取带红外线光斑的人脸照片；对带红外线光斑的人脸照片进行图像处理，从带红外线光斑的人脸照片中获取红外线光斑的形 态数据；根据红外线光斑的形态数据进行计算，获取目标人脸的防伪检测结果。下面对本实施例的实现细节进行具体的说明，以下内容仅为方便理解提供的实现细节，并非实施本方案的必须。
如图1所示，本申请第一实施例涉及一种人脸防伪检测方法，包括：
步骤101，向目标人脸投影红外线光斑，对目标人脸进行拍照，获取带红外线光斑的人脸照片。
在本实施例中，步骤101具体可以通过如点阵投影器等装置向目标人脸投影红外线光斑，并通过如红外镜头等装置对目标人脸进行拍照，获取带红外线光斑的人脸照片。
需要说明的是，本实施例不对步骤101向目标人脸投影红外线光斑的数量以及位置进行限定，在实际的使用过程中可以根据需求设置向目标人脸投影红外线光斑的数量以及位置，例如：可以设置向目标人脸的脸颊测投影一个红外线光斑，或者也可以如图2所示，设置向目标人脸的全脸均匀投影多个红外线光斑等，此处不对每种情况进行一一赘述。
步骤102，对带红外线光斑的人脸照片进行图像处理，从带红外线光斑的人脸照片中获取红外线光斑的形态数据。
在本实施例中，当带红外线光斑的人脸照片中包含多个红外线光斑时，步骤102可以获取全部红外线光斑的形态数据；也可以只获取部分红外线光斑的形态数据，具体地，步骤102可以对带红外线光斑的人脸照片进行图像处理，从多个红外线光斑中选取目标红外线光斑，获取目标红外线光斑的形态数据。
需要说明的是，当步骤102只获取部分红外线光斑的形态数据时，本专利申请实施例不对红外线光斑的获取个数和位置进行限定，步骤102可以根据 实际需要获取不同位置、不同数量的红外线光斑的形态数据。
通过对带红线光斑的人脸照片进行局部处理，从两个以上红外线光斑中选取目标红外线光斑，并根据目标红外线光斑的形态数据就可以完成人脸防伪检测，无需采用全局数据进行处理，减小了计算量，缩短了人脸防伪检测时间，并且，由于目标红外线光斑可以从两个以上红外线光斑中选取获得，使得本申请实施例可以适应不同场景下的不同需求，合理选取目标红外线光斑，从而进一步保证人脸防伪检测结果的准确性。
步骤103，根据红外线光斑的形态数据进行计算，获取目标人脸的防伪检测结果。
在本实施例中，步骤103具体可以通过两种方法实现：
一种方法是：当红外线光斑的形态数据为红外线光斑的像素值分布数据时，如图3所示，步骤103包括：
步骤301，从红外线光斑的像素值分布数据中，获取第一像素值衰减数据和第二像素值衰减数据。
在本实施例中，第一像素值衰减数据和第二像素值衰减数据为最大像素值数据衰减后得到的像素值，第一像素值衰减数据的衰减程度不同于第二像素值衰减数据的衰减程度。
具体地，步骤103可以通过如下步骤完成：
首先，根据红外线光斑的像素值分布数据，获取最大像素值数据。
为了便于理解，如图4所示，本实施例以红外线光斑的截面是横截面为例来显示红外线光斑的像素值分布情况。当然，在实际的使用过程中，还可以以纵截面或者其他角度的截面为例来显示红外线光斑的像素值分布情况，此处 不对每种情况进行一一赘述。
为了便于对后续步骤的理解，本实施例根据如图4所示的红外线光斑1的横截面对应的像素值分布数据4获取的最大像素值数据定义为I。
其次，根据最大像素值数据以及预先设置的第一像素值衰减规则，获取第一像素值衰减数据。
为了便于理解，在本实施例中第一像素值衰减规则可以预先设置为衰减
Figure PCTCN2019105449-appb-000001
则此时，第一像素值衰减数据为
Figure PCTCN2019105449-appb-000002
需要说明的是，以上仅为举例，本实施例并不对第一像素值衰减规则进行限定，在实际的使用过程中，可以根据需求设置第一像素值衰减规则，此处不做赘述。
然后，根据最大像素值数据以及预先设置的第二像素值衰减规则，获取第二像素值衰减数据。
具体地，本实施例中第二像素值衰减规则可以预先设置为衰减
Figure PCTCN2019105449-appb-000003
则此时，第二像素值衰减数据为
Figure PCTCN2019105449-appb-000004
需要说明的是，为了便于本领域技术人员能够充分理解本专利申请通过检测红外线光斑的形态的扩散程度检测真假人脸的原理，本实施例以第一像素值衰减数据的像素值大于第二像素值衰减数据的像素值为例进行说明，但是，可以理解的是，在实际的使用过程中，第一像素值衰减数据的像素值也可以小于第二像素值衰减数据的像素值，此时，可以通过检测红外线光斑的形态向中心点的像素值收敛程度，检测真假人脸。不管是检测红外线光斑的形态的扩散程度，还是检测检测红外线光斑的形态向中心点的光强收敛程度，原理上都是利用了红外线光斑的形态在不同介质上的形态变化不同进行的检测，属于相同 的原理和构思。
步骤302，获取第一像素值衰减数据在红外线光斑上对应的第一红外线子光斑的面积，以及第二像素值衰减数据在红外线光斑上对应的第二红外线子光斑的面积。
在本实施例中，如图4所示，第一像素值衰减数据在像素值分布数据4上的位置为5，第一像素值衰减数据在红外线光斑1上对应的第一红外线子光斑为3，步骤302可以通过计算获取第一红外线子光斑3的面积，在本实施例中，第一红外线子光斑3的面积用S3表示。
在本实施例中，如图4所示，第二像素值衰减数据在像素值分布数据4上的位置为6，第二像素值衰减数据在红外线光斑1上对应的第二红外线子光斑为2，步骤302可以通过计算获取第二红外线子光斑2的面积，在本实施例中，第二红外线子光斑2的面积用S2表示。
步骤303，将第一红外线子光斑的面积与第二红外线子光斑的面积之间的比例值与预先设置的真人人脸比例值区间进行比较，根据比较结果获取目标人脸的防伪检测结果。
在本实施例中，第一红外线子光斑的面积与第二红外线子光斑的面积之间的比例值R＝S2/S3。
在本实施例中，可以假定真人人脸比例值区间为[r1，r2]，当R∈[r1，r2]时，目标人脸为真人人脸，否则，目标人脸为假人人脸。
在本实施例中，假设通过如图3所示的步骤对真人人脸获取第一红外线子光斑的面积与第二红外线子光斑的面积之间的比例值为R1，对假人人脸获取的第一红外线子光斑的面积与第二红外线子光斑的面积之间的比例值为R2。
发明人经过反复实验获知，对于红外线散射能力比人体皮肤弱的材料，如不透明塑料、石膏或者纸张等，R1>R2；对于红外线散射能力比人体皮肤弱的材料，如半透明的塑料等，R1<R2。
如何合理地确定真人人脸比例值区间成为人脸防伪检测准确性的关键因素。经发明人反复实验发现，对于不同材料，第一红外线子光斑的面积与第二红外线子光斑的面积之间的比例值有明显差异，而对于同种材料第一红外线子光斑的面积与第二红外线子光斑的面积之间的比例值有固定的取值区间。
根据以上特点，在本实施例中，真人人脸比例值区间可以通过对不同的真人人脸采用如图3所示的步骤获取。当然，以上仅为举例，在实际的操作过程中，真人人脸的比例值区间还可以通过其他方法获取，此处不做一一赘述。
另外，需要说明的是，虽然，如图3所示的步骤中，仅以两个像素值数据——第一像素值衰减数据和第二像素值衰减数据，为例进行了说明，然而，本领域技术人员可以理解的是，在实际的使用过程中，也可以使用两个以上像数值数据进行计算，获取目标人脸的防伪检测结果。
具体地，例如：当存在三个像素值数据，其中，三个像数值数据分别为：第一像素值衰减数据、第二像素值衰减数据和第三像素值衰减数据，并且，第一像素值衰减数据的像素值大于第二像素值衰减数据的像素值，第二像素值衰减数据的像素值大于第三像数值衰减数据的像素值时，根据如图3所示的步骤分别获取：第二像素值衰减数据对应的第二红外线子光斑的面积与第一像素值衰减数据对应的第一红外线子光斑的面积之间的比例值R21；第三像素值衰减数据对应的第三红外线子光斑的面积与第二像素值衰减数据对应的第二红外线子光斑的面积之间的比例值R32；第三像素值衰减数据对应的第三红外线子光 斑的面积与第一像素值衰减数据对应的第一红外线子光斑的面积之间的比例值R31；对R21、R32和R31计算平均值，获得R321，当R321∈[r1，r2]时，目标人脸为真人人脸，否则，目标人脸为假人人脸。需要说明的是，本实施例对R21、R32和R31进行计算，根据计算结果检测人脸真伪的方法仅为举例，在实际的使用过程中还可以通过概率的方法检测人脸真伪，如：当R21、R32和R31有两个数值∈[r1，r2]时，目标人脸为真人人脸，否则，目标人脸为假人人脸。此处不对每种方法进行一一赘述。
另一种方法是：当红外线光斑的形态数据为通过主成分分析对红外光斑降维获得的特征向量时，如图5所示，步骤103包括：
步骤501，采用预先训练好的神经网络模型对特征向量进行计算，获得特征向量的标签值。
步骤502，将特征向量的标签值与预先设置的标签值阈值进行比较，根据比较结果获取目标人脸的防伪检测结果。
其中，神经网络模型是通过真人人脸的特征向量和假人人脸的特征向量训练获得的；标签值阈值是根据神经网络模型中真人人脸的特征向量的标签值以及假人人脸的特征向量的标签值获得的。
需要说明的是，在本实施例中，如图6所示，神经网络的训练方法可以包括：
步骤601，获取真人人脸对应的第一带红外线光斑的人脸照片集，以及假人人脸对应的第二带红外线光斑的人脸照片集。
具体地，向两个以上不同的真人人脸投射红外线光斑并拍照，获取两个以上不同的真人人脸对应的带红外线光斑的人脸照片，根据两个以上不同的真 人人脸对应的带红外线光斑的人脸照片，生成第一带红外线光斑的人脸照片集。
向两个以上不同的假人人脸投射红外线光斑并拍照，获取两个以上不同的假人人脸对应的带红外线光斑的人脸照片，根据两个以上不同的假人人脸对应的带红外线光斑的人脸照片，生成第二带红外线光斑的人脸照片集。
步骤602，采用主成分分析，分别获取第一带红外线光斑的人脸照片集对应的第一特征向量集，以及第二带红外线光斑的人脸照片集对应的第二特征向量集。
具体地，通过主成分分析，对第一对带红外线光斑的人脸照片集中每个带红外线光斑的人脸照片中的每个红外线光斑进行处理，获得对应的特征向量，根据该特征向量生成第一特征向量集。获取第二特征向量集的方法与获取第一特征向量集的方法相同，此处不再赘述。
步骤603，采用第一特征向量集和第二特征向量集对神经网络进行训练，获取训练好的神经网络模型。
需要说明的是，在如图6所示的技术方案中，并不对两个以上真人人脸和用两个以上假人人脸的数量进行限定，在实际的使用过程中，为了尽可能保证神经网络模型的可靠性，可以尽量多的获取真人人脸和假人人脸。
在本实施例中，可以假设神经网络模型中，真人人脸的特征向量的标签值为1，假人人脸的特征向量的标签值为0，则此时，标签值阈值可以为0.5。则，步骤502中，如果输入特征向量到训练好的神经网络模型后输出标签值大于0.5，则目标人脸为真人人脸，否则，目标人脸为假人人脸。需要说明的是，以上所述的标签值阈值为0.5仅为举例，在实际的使用过程中，为了提高人脸防伪检测的准确性，可以将标签值阈值设置的较大，如0.7、0.75等，为了避免 环境中干扰因素对人脸防伪检测结果的影响，也可以将标签值阈值设置的较小，如0.4、0.35等，此处不做赘述。
本申请实施例现对于现有技术而言，在向目标人脸投影红外线光斑并进行拍照，获取带红外线光斑的人脸照片，并根据人脸照片中红外线光斑的形态数据进行计算，获取目标人脸的防伪检测结果。本专利申请实施例利用了红外线光斑在不同介质上形态数据的差异性实现对人脸的防伪检测，其实现方法简单，并且便于用户操作。由于本专利申请是根据人脸照片中红外线光斑的形态数据进行计算，获取目标人脸的防伪检测结果，而人脸照片是在向目标人脸投影红外线光斑时进行拍照获得的，大大缩短了整个人脸防伪检测过程的时间，解决了现有技术需要根据人脸反射回的光线数据进行人脸防伪检测，造成检测时间过长的问题，也解决了现有技术需要用户主动做动作配合完成人脸防伪检测，动作过程缓慢，造成检测时间过长的问题，由于整个检测时间缩短了，进而提高了用户的使用体验。另外，由于本专利申请实施例利用了红外线光斑在不同介质上形态数据的差异性实现对人脸的防伪检测，使得本专利申请实施例只需要获取一帧人脸照片即可完成人脸的防伪检测，解决了现有技术需要用户主动做动作，获取多帧人脸照片完成人脸的防伪检测，造成用户操作复杂，并且人脸方位检测计算量大的问题。
如图7所示，本申请第二实施例涉及一种人脸防伪检测方法，该方法与如图1所示的基本相同，其区别在于，还包括：
步骤104，对带红外线光斑的人脸照片进行图像处理，从带红外线光斑的人脸照片中获取红外线光斑的位置坐标数据。
步骤105，根据红外线光斑的位置坐标数据和预先设置的标定数据，采 用三维重建算法进行计算，获取所述目标人脸的三维信息。
步骤106，根据目标人脸的三维信息判断目标人脸是否为人脸图形结构。
则此时，步骤102具体为步骤102’，如果目标人脸是人脸图形结构，对带红外线光斑的人脸照片进行图像处理，从带红外线光斑的人脸照片中获取红外线光斑的形态数据。
本申请实施例在达到图1所示的实施例所有有益效果基础上，先对人脸照片中是否存在人脸图形结构进行了判断，只有当存在人脸图形结构时才进一步对人脸真伪进行检测，进一步提升了人脸防伪检测的效率。
如图8所示，本申请第三实施例涉及一种人脸防伪检测装置，包括：
第一获取模块801，用于向目标人脸投影红外线光斑，对所述目标人脸进行拍照，获取带红外线光斑的人脸照片；
第二获取模块802，用于对所述第一获取模块801获取的所述带红外线光斑的人脸照片进行图像处理，从所述带红外线光斑的人脸照片中获取红外线光斑的形态数据；
检测模块803，用于根据所述第二获取模块802获取的所述红外线光斑的形态数据进行计算，获取所述目标人脸的防伪检测结果。
可选地，所述红外线光斑的形态数据为所述红外线光斑的像素值分布数据，如图9所示，所述检测模块803可以包括：
第一获取子模块901，用于从所述红外线光斑的像素值分布数据中，获取第一像素值衰减数据和第二像素值衰减数据；
第二获取子模块902，获取所述第一获取子模块获取的第一像素值衰减数据在所述红外线光斑上对应的第一红外线子光斑的面积，以及所述第一获 取子模块获取的第二像素值衰减数据在所述红外线光斑上对应的第二红外线子光斑的面积。
第一检测子模块903，用于将所述第二获取子模块902获取的所述第一红外线子光斑的面积与所述第二红外线子光斑的面积之间的比例值与预先设置的真人人脸比例值区间进行比较，根据比较结果获取所述目标人脸的防伪检测结果。
可选地，第一获取子模块901，具体用于根据所述红外线光斑的像素值分布数据，获取最大像素值数据；根据所述最大像素值数据以及预先设置的第一像素值衰减规则，获取第一像素值衰减数据；根据所述最大像素值数据以及预先设置的第二像素值衰减规则，获取第二像素值衰减数据。
可选地，所述红外线光斑的形态数据为通过主成分分析对所述红外光斑降维获得的特征向量，如图10所示，所述检测模块803可以包括：
第四获取子模块1001，用于采用预先训练好的神经网络模型对所述特征向量进行计算，获得特征向量的标签值；
第二检测子模块1002，用于将所述第四获取子模块1001获取的特征向量的标签值与预先设置的标签值阈值进行比较，根据比较结果获取所述目标人脸的防伪检测结果；
其中，所述神经网络模型是通过真人人脸的特征向量和假人人脸的特征向量训练获得的；
其中，所述标签值阈值是根据所述神经网络模型中真人人脸的特征向量的标签值和假人人脸的特征向量的标签值获得的。
可选地，如图10所示，所述检测模块803还可以包括：神经网络模 型训练子模块1003；
所述神经网络训练子模块1003可以包括：
第一获取单元10031，用于获取真人人脸对应的第一带红外线光斑的人脸照片集，以及假人人脸对应的第二带红外线光斑的人脸照片集；
第二获取单元10032，用于采用所述主成分分析，分别获取所述第一带红外线光斑的人脸照片集对应的第一特征向量集，以及所述第二带红外线光斑的人脸照片集对应的第二特征向量集；
训练单元10033，用于采用所述第一特征向量集和所述第二特征向量集对神经网络进行训练，获取训练好的神经网络模型。
可选地，如果所述带红外线光斑的人脸照片包括两个以上红外线光斑，第二获取模块802，还用于对所述带红线光斑的人脸照片进行图像处理，从所述两个以上红外线光斑中选取目标红外线光斑；获取所述目标红外线光斑的形态数据。
本实施例提供的人脸防伪检测装置的具体实现方法可以参见本申请第一实施例所述的人脸防伪检测方法，此处不再赘述。
本申请实施例现对于现有技术而言，在向目标人脸投影红外线光斑并进行拍照，获取带红外线光斑的人脸照片，并根据人脸照片中红外线光斑的形态数据进行计算，获取目标人脸的防伪检测结果。本专利申请实施例利用了红外线光斑在不同介质上形态数据的差异性实现对人脸的防伪检测，其实现方法简单，并且便于用户操作。由于本专利申请是根据人脸照片中红外线光斑的形态数据进行计算，获取目标人脸的防伪检测结果，而人脸照片是在向目标人脸投影红外线光斑时进行拍照获得的，大大缩短了整个人脸防伪检测过程的时间， 解决了现有技术需要根据人脸反射回的光线数据进行人脸防伪检测，造成检测时间过长的问题，也解决了现有技术需要用户主动做动作配合完成人脸防伪检测，动作过程缓慢，造成检测时间过长的问题，由于整个检测时间缩短了，进而提高了用户的使用体验。另外，由于本专利申请实施例利用了红外线光斑在不同介质上形态数据的差异性实现对人脸的防伪检测，使得本专利申请实施例只需要获取一帧人脸照片即可完成人脸的防伪检测，解决了现有技术需要用户主动做动作，获取多帧人脸照片完成人脸的防伪检测，造成用户操作复杂，并且人脸方位检测计算量大的问题。
如图11所示，本申请第四实施例涉及一种人脸防伪检测装置，该装置与如图8所示的基本相同，其区别在于，还包括：
第三获取模块804，用于对所述第一获取模块801获取的所述带红外线光斑的人脸照片进行图像处理，从所述带红外线光斑的人脸照片中获取红外线光斑的位置坐标数据；
第四获取模块805，用于根据所述第三获取模块804获取的所述红外线光斑的位置坐标数据和预先设置的标定数据，采用三维重建算法进行计算，获取所述目标人脸的三维信息；
判断模块806，用于根据所述第四获取模块805获取的所述目标人脸的三维信息判断所述目标人脸是否为人脸图形结构；
则第二获取模块802，具体用于如果所述判断模块806判断出所述目标人脸是人脸图形结构，对所述第一获取模块801获取的所述带红外线光斑的人脸照片进行图像处理，从所述带红外线光斑的人脸照片中获取红外线光斑的形态数据。
本实施例提供的人脸防伪检测装置的具体实现方法可以参见本申请第二实施例所述的人脸防伪检测方法，此处不再赘述。
本申请实施例在达到图8所示的实施例所有有益效果基础上，先对人脸照片中是否存在人脸图形结构进行了判断，只有当存在人脸图形结构时才进一步对人脸真伪进行检测，进一步提升了人脸防伪检测的效率。
本申请第五实施例涉及一种人脸防伪检测芯片，所述人脸防伪检测芯片与存储器通信连接，所述存储器存储有指令，所述指令被所述人脸防伪检测芯片执行，以使所述人脸防伪检测芯片能够执行以上本申请第一实施例和第二实施例所述的人脸防伪检测方法。
本申请实施例现对于现有技术而言，在向目标人脸投影红外线光斑并进行拍照，获取带红外线光斑的人脸照片，并根据人脸照片中红外线光斑的形态数据进行计算，获取目标人脸的防伪检测结果。本专利申请实施例利用了红外线光斑在不同介质上形态数据的差异性实现对人脸的防伪检测，其实现方法简单，并且便于用户操作。由于本专利申请是根据人脸照片中红外线光斑的形态数据进行计算，获取目标人脸的防伪检测结果，而人脸照片是在向目标人脸投影红外线光斑时进行拍照获得的，大大缩短了整个人脸防伪检测过程的时间，解决了现有技术需要根据人脸反射回的光线数据进行人脸防伪检测，造成检测时间过长的问题，也解决了现有技术需要用户主动做动作配合完成人脸防伪检测，动作过程缓慢，造成检测时间过长的问题，由于整个检测时间缩短了，进而提高了用户的使用体验。另外，由于本专利申请实施例利用了红外线光斑在不同介质上形态数据的差异性实现对人脸的防伪检测，使得本专利申请实施例只需要获取一帧人脸照片即可完成人脸的防伪检测，解决了现有技术需要用户 主动做动作，获取多帧人脸照片完成人脸的防伪检测，造成用户操作复杂，并且人脸方位检测计算量大的问题。
如图12所示，本申请第六实施例涉及一种电子设备，包括：存储器1201，以及以上第五实施例所述的人脸防伪检测芯片1202，所述存储器1201与所述人脸防伪检测芯片1202通信连接。
本申请实施例现对于现有技术而言，在向目标人脸投影红外线光斑并进行拍照，获取带红外线光斑的人脸照片，并根据人脸照片中红外线光斑的形态数据进行计算，获取目标人脸的防伪检测结果。本专利申请实施例利用了红外线光斑在不同介质上形态数据的差异性实现对人脸的防伪检测，其实现方法简单，并且便于用户操作。由于本专利申请是根据人脸照片中红外线光斑的形态数据进行计算，获取目标人脸的防伪检测结果，而人脸照片是在向目标人脸投影红外线光斑时进行拍照获得的，大大缩短了整个人脸防伪检测过程的时间，解决了现有技术需要根据人脸反射回的光线数据进行人脸防伪检测，造成检测时间过长的问题，也解决了现有技术需要用户主动做动作配合完成人脸防伪检测，动作过程缓慢，造成检测时间过长的问题，由于整个检测时间缩短了，进而提高了用户的使用体验。另外，由于本专利申请实施例利用了红外线光斑在不同介质上形态数据的差异性实现对人脸的防伪检测，使得本专利申请实施例只需要获取一帧人脸照片即可完成人脸的防伪检测，解决了现有技术需要用户主动做动作，获取多帧人脸照片完成人脸的防伪检测，造成用户操作复杂，并且人脸方位检测计算量大的问题。
本申请第七实施例涉及一种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上本申请第一实施例和第二实施例所述 的人脸防伪检测方法。
本申请实施例现对于现有技术而言，在向目标人脸投影红外线光斑并进行拍照，获取带红外线光斑的人脸照片，并根据人脸照片中红外线光斑的形态数据进行计算，获取目标人脸的防伪检测结果。本专利申请实施例利用了红外线光斑在不同介质上形态数据的差异性实现对人脸的防伪检测，其实现方法简单，并且便于用户操作。由于本专利申请是根据人脸照片中红外线光斑的形态数据进行计算，获取目标人脸的防伪检测结果，而人脸照片是在向目标人脸投影红外线光斑时进行拍照获得的，大大缩短了整个人脸防伪检测过程的时间，解决了现有技术需要根据人脸反射回的光线数据进行人脸防伪检测，造成检测时间过长的问题，也解决了现有技术需要用户主动做动作配合完成人脸防伪检测，动作过程缓慢，造成检测时间过长的问题，由于整个检测时间缩短了，进而提高了用户的使用体验。另外，由于本专利申请实施例利用了红外线光斑在不同介质上形态数据的差异性实现对人脸的防伪检测，使得本专利申请实施例只需要获取一帧人脸照片即可完成人脸的防伪检测，解决了现有技术需要用户主动做动作，获取多帧人脸照片完成人脸的防伪检测，造成用户操作复杂，并且人脸方位检测计算量大的问题。
本领域的普通技术人员可以理解，上述各实施例是实现本申请的具体实施例，而在实际应用中，可以在形式上和细节上对其作各种改变，而不偏离本申请的精神和范围。

Claims (11)

  1. 一种人脸防伪检测方法，其特征在于，包括：
    向目标人脸投影红外线光斑，对所述目标人脸进行拍照，获取带红外线光斑的人脸照片；
    对所述带红外线光斑的人脸照片进行图像处理，从所述带红外线光斑的人脸照片中获取红外线光斑的形态数据；
    根据所述红外线光斑的形态数据进行计算，获取所述目标人脸的防伪检测结果。
  2. 根据权利要求1所述的方法，其特征在于，所述红外线光斑的形态数据为所述红外线光斑的像素值分布数据；
    所述根据所述红外线光斑的形态数据进行计算，获取所述目标人脸的防伪检测结果包括：
    从所述红外线光斑的像素值分布数据中，获取第一像素值衰减数据和第二像素值衰减数据，其中，所述第一像素值衰减数据和所述第二像素值衰减数据为最大像素值数据衰减后得到的像素值，所述第一像素值衰减数据的衰减程度不同于所述第二像素值衰减数据的衰减程度；
    获取所述第一像素值衰减数据在所述红外线光斑上对应的第一红外线子光斑的面积，以及所述第二像素值衰减数据在所述红外线光斑上对应的第二红外线子光斑的面积；
    将所述第一红外线子光斑的面积与所述第二红外线子光斑的面积之间的比例值与预先设置的真人人脸比例值区间进行比较，根据比较结果获取所述目标人脸的防伪检测结果。
  3. 根据权利要求2所述的方法，其特征在于，所述从所述红外线光斑的像素值分布数据中，获取第一像素值衰减数据和第二像素值衰减数据包括：
    根据所述红外线光斑的像素值分布数据，获取所述最大像素值数据；
    根据所述最大像素值数据以及预先设置的第一像素值衰减规则，获取第一像素值衰减数据；
    根据所述最大像素值数据以及预先设置的第二像素值衰减规则，获取第二像素值衰减数据。
  4. 根据权利要求2所述的方法，其特征在于，所述第一像素值衰减数据的像素值大于所述第二像素值衰减数据的像素值。
  5. 根据权利要求1所述的方法，其特征在于，所述红外线光斑的形态数据为通过主成分分析对所述红外光斑降维获得的特征向量；
    所述根据所述红外线光斑的形态数据进行计算，获取所述目标人脸的防伪检测结果包括：
    采用预先训练好的神经网络模型对所述特征向量进行计算，获得特征向量的标签值；
    将所述特征向量的标签值与预先设置的标签值阈值进行比较，根据比较结果获取所述目标人脸的防伪检测结果；
    其中，所述神经网络模型是通过真人人脸的特征向量和假人人脸的特征向量训练获得的，所述标签值阈值是根据所述神经网络模型中真人人脸的特征向量的标签值和假人人脸的特征向量的标签值获得的。
  6. 根据权利要求5所述的方法，其特征在于，所述神经网络模型的训练方法包括：
    获取真人人脸对应的第一带红外线光斑的人脸照片集，以及假人人脸对应的第二带红外线光斑的人脸照片集；
    采用所述主成分分析，分别获取所述第一带红外线光斑的人脸照片集对应的第一特征向量集，以及所述第二带红外线光斑的人脸照片集对应的第二特征向量集；
    采用所述第一特征向量集和所述第二特征向量集对神经网络进行训练，获取训练好的神经网络模型。
  7. 根据权利要求1所述的方法，其特征在于，所述带红外线光斑的人脸照片包括两个以上红外线光斑；所述对所述带红外线光斑的人脸照片进行图像处理，从所述带红外线光斑的人脸照片中获取红外线光斑的形态数据，包括：
    对所述带红线光斑的人脸照片进行图像处理，从所述两个以上红外线光斑中选取目标红外线光斑；
    获取所述目标红外线光斑的形态数据。
  8. 根据权利要求1至7中任意一项所述的方法，其特征在于，还包括：
    对所述带红外线光斑的人脸照片进行图像处理，从所述带红外线光斑的人脸照片中获取红外线光斑的位置坐标数据；
    根据所述红外线光斑的位置坐标数据和预先设置的标定数据，采用三维重建算法进行计算，获取所述目标人脸的三维信息；
    根据所述目标人脸的三维信息判断所述目标人脸是否为人脸图形结构；
    所述对所述带红外线光斑的人脸照片进行图像处理，从所述带红外线光斑的人脸照片中获取红外线光斑的形态数据具体为：
    如果所述目标人脸是人脸图形结构，对所述带红外线光斑的人脸照片进行图像处理，从所述带红外线光斑的人脸照片中获取红外线光斑的形态数据。
  9. 一种人脸防伪检测芯片，其特征在于，所述人脸防伪检测芯片与存储器通信连接，所述存储器存储有指令，所述指令被所述人脸防伪检测芯片执行，以使所述人脸防伪检测芯片能够执行如权利要求1至8中任一所述的人脸防伪检测方法。
  10. 一种电子设备，其特征在于，包括：存储器，以及如权利要求9所述的人脸防伪检测芯片，所述存储器与所述人脸防伪检测芯片通信连接。
  11. 一种计算机可读存储介质，存储有计算机程序，其特征在于，所述计算机程序被处理器执行时实现权利要求1至8中任一项所述的人脸防伪检测方法。
PCT/CN2019/105449 2019-09-11 2019-09-11 人脸防伪检测方法、装置、芯片、电子设备和计算机可读介质 WO2021046773A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2019/105449 WO2021046773A1 (zh) 2019-09-11 2019-09-11 人脸防伪检测方法、装置、芯片、电子设备和计算机可读介质
CN201980001918.9A CN110720105A (zh) 2019-09-11 2019-09-11 人脸防伪检测方法、装置、芯片、电子设备和计算机可读介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2019/105449 WO2021046773A1 (zh) 2019-09-11 2019-09-11 人脸防伪检测方法、装置、芯片、电子设备和计算机可读介质

Publications (1)

Publication Number Publication Date
WO2021046773A1 true WO2021046773A1 (zh) 2021-03-18

Family

ID=69216606

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/105449 WO2021046773A1 (zh) 2019-09-11 2019-09-11 人脸防伪检测方法、装置、芯片、电子设备和计算机可读介质

Country Status (2)

Country Link
CN (1) CN110720105A (zh)
WO (1) WO2021046773A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113095284A (zh) * 2021-04-30 2021-07-09 平安国际智慧城市科技股份有限公司 人脸选取方法、装置、设备及计算机可读存储介质

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113673287B (zh) * 2020-05-15 2023-09-12 深圳市光鉴科技有限公司 基于目标时间节点的深度重建方法、系统、设备及介质
CN114676403A (zh) * 2020-12-24 2022-06-28 华为云计算技术有限公司 认证系统、设备及方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105637532A (zh) * 2015-06-08 2016-06-01 北京旷视科技有限公司 活体检测方法、活体检测系统以及计算机程序产品
CN107545241A (zh) * 2017-07-19 2018-01-05 百度在线网络技术（北京）有限公司 神经网络模型训练及活体检测方法、装置及存储介质
CN108537152A (zh) * 2018-03-27 2018-09-14 百度在线网络技术（北京）有限公司 用于检测活体的方法和装置
US20190095701A1 (en) * 2017-09-27 2019-03-28 Lenovo (Beijing) Co., Ltd. Living-body detection method, device and storage medium
CN110197132A (zh) * 2019-05-13 2019-09-03 径卫视觉科技（上海）有限公司 一种用于识别活体人脸的方法以及相应的系统

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7454046B2 (en) * 2005-09-20 2008-11-18 Brightex Bio-Photonics, Llc Method and system for analyzing skin conditions using digital images
CN102622588B (zh) * 2012-03-08 2013-10-09 无锡中科奥森科技有限公司 双验证人脸防伪方法及装置
JP2015064311A (ja) * 2013-09-26 2015-04-09 株式会社日立ハイテクノロジーズ 赤外線検査装置および赤外線検査方法
US9396537B2 (en) * 2014-09-09 2016-07-19 EyeVerify, Inc. Systems and methods for liveness analysis
CN111985294A (zh) * 2015-09-01 2020-11-24 北京上古视觉科技有限公司 一种具有活体检测功能的虹膜识别系统
CN105184277B (zh) * 2015-09-29 2020-02-21 杨晴虹 活体人脸识别方法以及装置
CN106407914B (zh) * 2016-08-31 2019-12-10 北京旷视科技有限公司 用于检测人脸的方法、装置和远程柜员机系统
CN107590430A (zh) * 2017-07-26 2018-01-16 百度在线网络技术（北京）有限公司 活体检测方法、装置、设备及存储介质
CN109923555B (zh) * 2019-01-29 2020-11-27 深圳市汇顶科技股份有限公司 指纹检测方法、指纹检测装置和电子设备

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105637532A (zh) * 2015-06-08 2016-06-01 北京旷视科技有限公司 活体检测方法、活体检测系统以及计算机程序产品
CN107545241A (zh) * 2017-07-19 2018-01-05 百度在线网络技术（北京）有限公司 神经网络模型训练及活体检测方法、装置及存储介质
US20190095701A1 (en) * 2017-09-27 2019-03-28 Lenovo (Beijing) Co., Ltd. Living-body detection method, device and storage medium
CN108537152A (zh) * 2018-03-27 2018-09-14 百度在线网络技术（北京）有限公司 用于检测活体的方法和装置
CN110197132A (zh) * 2019-05-13 2019-09-03 径卫视觉科技（上海）有限公司 一种用于识别活体人脸的方法以及相应的系统

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113095284A (zh) * 2021-04-30 2021-07-09 平安国际智慧城市科技股份有限公司 人脸选取方法、装置、设备及计算机可读存储介质

Also Published As

Publication number Publication date
CN110720105A (zh) 2020-01-21

Similar Documents

Publication Publication Date Title
US10956714B2 (en) Method and apparatus for detecting living body, electronic device, and storage medium
CN108764091B (zh) 活体检测方法及装置、电子设备和存储介质
US11010967B2 (en) Three dimensional content generating apparatus and three dimensional content generating method thereof
US10198623B2 (en) Three-dimensional facial recognition method and system
CN110383288B (zh) 人脸识别的方法、装置和电子设备
CN105335722B (zh) 一种基于深度图像信息的检测系统及方法
CN108764071B (zh) 一种基于红外和可见光图像的真实人脸检测方法及装置
EP3576017A1 (en) Method, apparatus, and device for determining pose of object in image, and storage medium
CN105740775B (zh) 一种三维人脸活体识别方法及其装置
US8750573B2 (en) Hand gesture detection
US8792722B2 (en) Hand gesture detection
US7925093B2 (en) Image recognition apparatus
CN112052831B (zh) 人脸检测的方法、装置和计算机存储介质
WO2021046773A1 (zh) 人脸防伪检测方法、装置、芯片、电子设备和计算机可读介质
JP2021522591A (ja) 三次元実物体を実物体の二次元のスプーフと区別するための方法
CN112036339B (zh) 人脸检测的方法、装置和电子设备
WO2020243969A1 (zh) 人脸识别的装置、方法和电子设备
CN111680574B (zh) 一种人脸检测方法、装置、电子设备和存储介质
US20160245641A1 (en) Projection transformations for depth estimation
CN111598065A (zh) 深度图像获取方法及活体识别方法、设备、电路和介质
CN113128428B (zh) 基于深度图预测的活体检测方法和相关设备
CN111079470A (zh) 人脸活体检测的方法和装置
JP2017084065A (ja) なりすまし検出装置
CN112153300A (zh) 多目摄像头曝光方法、装置、设备及介质
US11205064B1 (en) Measuring quality of depth images in real time

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: 19944975

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: 19944975

Country of ref document: EP

Kind code of ref document: A1