CN112183357B - Multi-scale living body detection method and system based on deep learning - Google Patents
Multi-scale living body detection method and system based on deep learning Download PDFInfo
- Publication number
- CN112183357B CN112183357B CN202011047776.0A CN202011047776A CN112183357B CN 112183357 B CN112183357 B CN 112183357B CN 202011047776 A CN202011047776 A CN 202011047776A CN 112183357 B CN112183357 B CN 112183357B
- Authority
- CN
- China
- Prior art keywords
- scale
- image
- living body
- information
- features
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 81
- 238000013135 deep learning Methods 0.000 title claims abstract description 36
- 230000004927 fusion Effects 0.000 claims abstract description 35
- 230000006399 behavior Effects 0.000 claims abstract description 25
- 238000000034 method Methods 0.000 claims abstract description 22
- 238000003384 imaging method Methods 0.000 claims abstract description 12
- 238000013136 deep learning model Methods 0.000 claims abstract description 5
- 238000013527 convolutional neural network Methods 0.000 claims description 22
- 230000003044 adaptive effect Effects 0.000 claims description 12
- 238000000605 extraction Methods 0.000 claims description 11
- 230000007613 environmental effect Effects 0.000 claims description 5
- 230000001815 facial effect Effects 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000012984 biological imaging Methods 0.000 claims description 3
- 238000001574 biopsy Methods 0.000 claims 2
- 238000001727 in vivo Methods 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000691 measurement method Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Human Computer Interaction (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The multi-scale living body detection method based on deep learning comprises the following steps: inputting a picture and extracting a multi-scale image; extracting multi-scale features from the multi-scale image: extracting multi-scale features from the multi-scale image by using a deep learning model to obtain face image information features, environment information features and behavior information features; step three, acquiring multi-scale fusion characteristics: performing feature fusion on the extracted multi-scale features by adopting different constraints to obtain multi-scale fusion features; inputting the multiscale fusion characteristics into a classification network, outputting a living body score, and obtaining a living body detection result according to a threshold value. Compared with the current living body detection method based on a single face area, the method provided by the invention has better scene adaptability and higher detection accuracy in a worse imaging environment.
Description
Technical Field
The invention belongs to the technical field of image recognition, and particularly relates to a multi-scale living body detection method and system based on deep learning.
Background
Biometric identification, particularly facial identification, has long been a research hotspot in the field of computer vision. With the development of deep learning and the promotion of hardware operation equipment, the face recognition technology is widely applied to various fields, such as mobile phone face unlocking, access control robot face attendance checking and online face recognition payment. The prior face recognition technology has potential safety hazards that identity information is stolen, and lawbreakers can carry out identity verification through fake living face information and carry out illegal activities such as stealing property, endangering public security and the like after the identity verification. Face recognition applications require a living detection method to identify whether a given face is living or not, where living is living with living objects, and typical non-living attacks include printed face paper attacks, video face attacks, and 3D face mask attacks. There are many RGB image face living body measurement methods based on conventional vision and deep learning, however, a single living body measurement method based on face information is easily affected by environment and equipment, such as illumination environment and equipment imaging quality, and in some poor environments, it is difficult for the algorithm to distinguish living bodies from non-living bodies.
Disclosure of Invention
According to the multi-scale living body detection method and system based on deep learning, starting from the multi-scale feature method, the feature extraction neural network is used for extracting the face features, the environmental information features and the behavior information features near the face under multiple scales and carrying out feature fusion, so that the adaptability of an algorithm under different environments is enhanced, and the detection accuracy under a poor imaging environment is improved.
The technical scheme provided by the invention is as follows:
according to one aspect of the present invention, there is provided a multi-scale living body detection method based on deep learning, comprising the steps of: inputting a picture and extracting a multi-scale image; extracting multi-scale features from the multi-scale image: extracting multi-scale features from the multi-scale image by using a deep learning model to obtain face image information features, environment information features and behavior information features; step three, acquiring multi-scale fusion characteristics: performing feature fusion on the extracted multi-scale features by adopting different constraints to obtain multi-scale fusion features; inputting the multiscale fusion characteristics into a classification network, outputting a living body score, and obtaining a living body detection result according to a threshold value.
Preferably, in the above-mentioned multi-scale living body detection method based on deep learning, in step one, a multi-scale image is extracted from a target to be detected in an input picture, the multi-scale image includes a low-scale face image information image, a medium-scale environment information image, and a high-scale behavior information image, and the multi-scale image is an RGB image.
Preferably, in the above-mentioned multi-scale living body detection method based on deep learning, in the third step, the multi-scale fusion feature may be expressed by formula (2):
L(G l ,G m ,G h )＝λ 1 F l +λ 2 F m +λ 3 F h (2)
in formula (2), G l ,G m ,G h Respectively acquired low, medium and high scale images, F l Lambda for extracted low-scale imaging features 1 For its constraint, F m Lambda is the extracted mesoscale environmental information feature 2 For its constraint, F h Lambda is the extracted high-scale behavior information feature 3 Is constrained by it.
According to another aspect of the present invention, there is provided a multi-scale living body detection system based on deep learning, using the multi-scale living body detection method based on deep learning, the multi-scale living body detection system including an adaptive multi-scale image acquisition module, a convolutional neural network creation module, and a multi-scale living body detection module, wherein the convolutional neural network creation module is configured to design a convolutional neural network model, perform living body judgment on an input target to be detected, and output a living body detection score; the multi-scale living body detection module is used for extracting multi-scale image information of the target to be detected, inputting the multi-scale image information into the convolutional neural network model, and fusing multi-scale living body detection scores output by the convolutional neural network model to obtain a living body detection result of the target to be detected.
Preferably, in the above-mentioned multi-scale living body detection system based on deep learning, the adaptive multi-scale image acquisition module includes: the low-scale biological information acquisition unit is used for acquiring a facial information image according to the position information of the target to be detected;
the mesoscale environment information acquisition unit is used for acquiring an image containing an environment background by using an adaptive method independent of the resolution of a camera and the size of the image according to the position information of the target to be detected; the high-scale behavior information acquisition unit is used for acquiring an image containing the behavior information of the target to be detected by using an adaptive method independent of the resolution of the camera and the size of the image according to the position information of the target to be detected.
Preferably, in the above multi-scale living body detection system based on deep learning, the convolutional neural network creation module includes: the RGB image characteristic information extraction network is used for constructing a multi-level deep neural network and extracting multi-level semantic characteristic information of the object to be detected under different scales; the RGB image characteristic information classification network is used for constructing a multi-level semantic characteristic information fusion network, fusing the extracted semantic information of the object to be detected and outputting a living body score, wherein the score is between 0 and 1, if the object is living, the network output result is 1, and if the object is non-living, the network output result is 0.
Preferably, in the above-mentioned deep learning-based multi-scale living body detection system, the multi-scale living body detection module includes: a low-scale biological imaging feature constraint unit for giving constraint weight to the facial imaging features extracted from the low-scale image; the mesoscale environment information feature constraint unit is used for giving constraint weight to the environment features extracted from the mesoscale image; the high-scale behavior information feature constraint unit is used for giving constraint weights to the behavior features of the target to be detected extracted from the high-scale image.
Compared with the prior art, the invention has the beneficial effects that:
by utilizing the technical scheme provided by the invention, the living body detection is carried out in the face identification process, the multi-scale characteristics and the characteristic fusion method are adopted, and the multi-scale visual information is utilized, so that compared with the living body detection method which only focuses on the face area at present, the environment information characteristics and the behavior information characteristics are additionally introduced and fused, and then the fused characteristics are input into the classification network to carry out living body scoring on the target to be detected, thereby realizing living body detection. Compared with the current living body detection method based on a single face area, the method provided by the invention has better scene adaptability and higher detection accuracy in a worse imaging environment.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below.
FIG. 1 is a flow chart of a deep learning based multi-scale in-vivo detection method of the present invention;
FIG. 2 is a network structure diagram of the deep learning based multi-scale in-vivo detection system of the present invention;
fig. 3 is a multi-scale image extraction structure diagram of the multi-scale living body detection method based on deep learning of the present invention.
Detailed Description
The multi-scale living body detection method based on deep learning adopts a deep learning framework, designs a multi-scale fusion characteristic method and completes living body detection on the basis.
The principle of the invention is as follows: 1. ) The in-vivo detection problem is expressed as a multi-scale feature detection model, each scale focusing on a different visual feature. The imaging features of faces are of interest at low scales, i.e. imaging features under different media (paper, screen and mask), such as mole marks of screen secondary imaging and abnormal faces in paper attack. The mesoscale focuses on environmental features of the area near the face, such as paper boundaries, screen boundaries. The high scale focuses on behavior information of the object to be measured, such as hand movements. 2. ) The method is characterized in that features of different scales are fused, namely, attention mechanisms (different attention degrees are adopted) are used for different types of feature information, so that the method has adaptability in different scenes.
The multi-scale living body detection method based on deep learning comprises three parts: carrying out multi-scale image extraction on a target to be detected in an input picture; extracting features of images with different scales by using a deep learning model; fusing the multi-scale features; and scoring the fusion characteristics by using a classification network, and judging whether the target to be detected is a living body or not through a threshold value. As shown in fig. 1, in the multi-scale living body detection method based on deep learning of the present invention, from picture input to living body detection result output includes the following steps:
inputting a picture, and extracting a multi-scale image s1: and carrying out target detection on the input image, acquiring the position information of the target to be detected, and carrying out self-adaptive multi-scale image information acquisition according to the position information of the target to be detected. The multi-scale image comprises a low-scale face image information image, a medium-scale environment information image and a high-scale behavior information image, wherein the target detection method is not limited to the traditional computer vision and deep learning method.
Extracting multi-scale features s2 from the multi-scale image, namely extracting multi-scale features of the multi-scale image by using a deep learning model to obtain multi-scale features of the object to be detected under different scales, wherein the multi-scale features comprise face image information features, environment information features and behavior information features;
thirdly, acquiring a multi-scale fusion feature s3, namely performing feature fusion on the extracted multi-scale feature by adopting different feature constraints to obtain the multi-scale fusion feature;
inputting the multi-scale fusion features into a classification network, outputting a living body score, and obtaining a living body detection result s4 according to a threshold value, namely inputting the fusion features into the classification network to obtain a living body detection score, and obtaining the living body detection result according to a set score threshold value.
The following describes the specific implementation steps of the deep learning-based multi-scale living body detection method according to the present invention with reference to fig. 2 and 3, and the overall operation flow is as follows:
inputting a picture, extracting multiple scalesImage s1, given an RGB image 12 (FIG. 3) acquired by the device, this image is denoted G R The width and height of the image are denoted w and h, respectively. Face position information (x) is acquired by a general face detection method (the detection method is not limited to the conventional vision method or the deep learning method) l ，y l ，w l ，h l ) Wherein x is l ，y l Is the central position coordinate of the face region, w l And h l The face acquisition region is a low-scale image acquisition region 15 (see fig. 3) for the width and height of the face region. Based on the width and height of the image and the position information of the face region, a self-adaptive computing method is used to obtain a mesoscale image acquisition region 14 and a high-scale image acquisition region 13 (as shown in fig. 3), wherein the computing method is shown in formula (1):
wherein S is i1 And S is i2 The scale parameters are adjustable according to the equipment and the actual scene. Wherein the acquired low-scale image is marked as G l The acquired mesoscale image is denoted as G m And the acquired high-scale image is noted.
And secondly, extracting multi-scale features s2 from the images with different scales, and inputting the images with different scales, such as a high-scale image 1, a middle-scale image 2 and a low-scale image 3 in the figure 2, into the respective corresponding feature extraction convolutional neural network in a feature extraction stage to obtain the corresponding multi-scale features. As shown in fig. 2, a high-scale image 1 is input into a high-scale image feature extraction convolutional neural network 4 to obtain high-scale features 7; inputting the mesoscale image 2 into a mesoscale image feature extraction convolutional neural network 5 to obtain mesoscale features 8; and the low-scale image 3 is input into a low-scale image feature extraction convolutional neural network 6 to obtain low-scale features 9.
And thirdly, obtaining a multi-scale fusion feature s3. And carrying out feature fusion on the acquired multi-scale features in the multi-scale fusion feature unit 10 by adopting different constraints to obtain multi-scale fusion features. Wherein the fusion features are noted asL(G l 、G m 、G h ) Wherein G is l 、G m And G h Corresponding features are respectively marked as F l 、F m And F h The expression is shown as (2):
L(G l ，G m ，G h )＝λ 1 F l +λ 2 F m +λ 3 F h (2)
in formula (2), F l Lambda is the extracted low-scale face image feature 1 Is constrained by; f (F) m Lambda is the extracted mesoscale environmental information feature 2 Is constrained by; f (F) h Lambda is the extracted high-scale behavior information feature 3 Is constrained by it. And acquiring fusion features by using a weighted average method, and taking different attention degrees for features of different scales. Wherein lambda is 1 、λ 2 And lambda (lambda) 3 And the device is not fixed, and can be correspondingly adjusted according to the environment and equipment in the actual scene.
And fourthly, inputting the fusion characteristics into a classification network, outputting a living body score, and obtaining a living body detection result s4 according to a threshold value. The fusion feature L is input into the classification network module 11, and the output result is constrained to [0,1] by using a Sigmoid function, so that a probability value P of the output result is obtained, namely the score of the living body detection of the target to be detected. According to the set living body detection threshold value α, living bodies are classified as living bodies for which the threshold value α is greater than or equal to, and non-living bodies are classified as non-living bodies for which the threshold value α is less than.
A deep learning based multi-scale in vivo detection system comprising: the system comprises an adaptive multi-scale image acquisition module, a convolutional neural network creation module and a multi-scale living body detection module. Wherein,
an adaptive multi-scale image acquisition module comprising: the low-scale biological information acquisition unit is used for acquiring face image information according to the position information of the target to be detected; the mesoscale environment information acquisition unit is used for acquiring an image containing environment information by using an adaptive method independent of the resolution of a camera and the size of the image according to the position information of the target to be detected; the high-scale behavior information acquisition unit is used for acquiring an image containing behavior information by using an adaptive method independent of the resolution of a camera and the size of the image according to the position information of the target to be detected.
The convolutional neural network creation module is used for designing a convolutional neural network model, performing living body judgment on an input target to be detected and outputting a living body detection score, and comprises the following steps: the RGB image characteristic information extraction network is used for constructing a multi-level deep neural network for extracting multi-scale characteristics of the object to be detected under different scales; the RGB image characteristic information classification network is used for constructing a multi-level semantic characteristic information fusion network, fusing the extracted semantic information of the object to be detected and outputting a living body score, wherein the score is between 0 and 1, if the object is living, the network output result is 1, and if the object is non-living, the network output result is 0.
The multi-scale living body detection module is used for extracting multi-scale image information of the target to be detected, inputting the multi-scale image information into the convolutional neural network module, and fusing the multi-scale living body detection scores output by the model to obtain a living body detection result of the target to be detected. Comprising the following steps: the low-scale biological imaging feature constraint unit is used for giving constraint weight to the face imaging information features extracted from the low-scale image; the mesoscale environment information feature constraint unit is used for giving constraint weight to the environment information features extracted from the mesoscale image; the high-scale behavior information feature constraint unit is used for giving constraint weights to the behavior information features of the target to be detected extracted from the high-scale image; and the multi-scale fusion characteristic unit is used for fusing multi-scale constraint characteristics, inputting the multi-scale constraint characteristics into the classification network and outputting a living body detection result. Wherein the multiscale fusion feature is as described in equation (2) above.
The method of the invention is trained and evaluated in the living body detection data set Siw, celebA_Spoof and the data of the self-acquisition application scene, and the evaluation method adopts FAR (False Aceeptance Rate) and FRR (False Rejection Rate). At a threshold of 0.5, the method provided by the invention is superior to a deep learning living body detection method based on a single scale.
At lambda 1 、λ 2 、λ 3 When the values are respectively 0.5, 0.3 and 0.2 and the threshold value is set to be 0.5,the test results FAR and FRR are shown in the following table. Wherein FAR represents the proportion of the false face judged as the true face, and FRR represents the proportion of the true face judged as the false face. It can be seen from the table that the multi-scale approach is superior to the single-scale approach in both FAR and FRR.
| FAR(％) | FRR(％) | |
| Single scale method | 1.0 | 5.0 |
| Multi-scale method | 0.8 | 4.2 |
The invention discloses a multi-scale living body detection method based on deep learning, which adopts a convolutional neural network in the deep learning, designs a multi-scale human face living body detection method based on background semantic information, images a target object to be detected, extracts multi-scale information, inputs the multi-scale information into a convolutional neural network model to obtain detection confidence, fuses the detection confidence under the multi-scale, and determines whether the target object to be detected is a living body. Compared with a living body detection method based on a single face, the method has better detection accuracy.
The above examples are only specific embodiments of the present invention for illustrating the technical solution of the present invention, but not for limiting the scope of the present invention, and although the present invention has been described in detail with reference to the foregoing examples, it will be understood by those skilled in the art that the present invention is not limited thereto: any person skilled in the art may modify or easily conceive of the technical solution described in the foregoing embodiments, or perform equivalent substitution of some of the technical features, while remaining within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention, and are intended to be included in the scope of the present invention. Therefore, the protection scope of the invention is subject to the protection scope of the claims.
Claims (6)
1. A multi-scale living body detection method based on deep learning, which is characterized by comprising the following steps:
inputting a picture, extracting a multi-scale image, giving an RGB image acquired by a piece of equipment, marking the image as GR, marking the width and height of the image as w and h respectively, and acquiring face position information (x) by a general face detection method l ，y l ，w l ，h l ) Wherein x is l ，y l Is the central position coordinate of the face region, w l And h l For the width and the height of the face area, according to the width and the height of the image and the position information of the face area, a self-adaptive computing method is used for acquiring a mesoscale image acquisition area and a high-scale image acquisition area, and the computing method is shown as a formula (1):
wherein s is i1 Sum s i2 The scale parameters can be adjusted according to equipment and actual scenes, wherein the acquired low-scale image is marked as G l The acquired mesoscale image is denoted as G m The acquired high-scale image is denoted as G h ；
Extracting multi-scale features from the multi-scale image: extracting multi-scale features from the multi-scale image by using a deep learning model to obtain face image information features, environment information features and behavior information features;
step three, acquiring multi-scale fusion characteristics: performing feature fusion on the extracted multi-scale features by adopting different constraints to obtain multi-scale fusion features,
the multiscale fusion feature can be represented by formula (2):
L(G l ,G m ,G h )＝λ 1 F l +λ 2 F m +λ 3 F h (2)
in formula (2), G l ,G m ,G h Respectively acquired low, medium and high scale images, F l Lambda for extracted low-scale imaging features 1 For its constraint, F m Lambda is the extracted mesoscale environmental information feature 2 For its constraint, F h Lambda is the extracted high-scale behavior information feature 3 Is constrained by;
and step four, inputting the multi-scale fusion characteristics into a classification network, outputting a living body score, and obtaining a living body detection result according to a threshold value.
2. The deep learning-based multi-scale living body detection method according to claim 1, wherein in the first step, a multi-scale image is extracted for a target to be detected in an input picture, the multi-scale image including a low-scale face image information image, a middle-scale environment information image, and a high-scale behavior information image, the multi-scale image being an RGB image.
3. A multi-scale living body detection system based on deep learning, which uses the multi-scale living body detection method based on deep learning as set forth in claim 1 or 2, characterized in that the multi-scale living body detection system comprises an adaptive multi-scale image acquisition module, a convolutional neural network creation module, and a multi-scale living body detection module, wherein,
the convolutional neural network creation module is used for designing a convolutional neural network model, performing living judgment on an input target to be detected and outputting a living detection score;
the multi-scale living body detection module is used for extracting multi-scale image information of a target to be detected, inputting the multi-scale image information into the convolutional neural network model, and fusing multi-scale living body detection scores output by the convolutional neural network model to obtain a living body detection result of the target to be detected.
4. The deep learning based multi-scale living detection system of claim 3, wherein the adaptive multi-scale image acquisition module includes:
the low-scale biological information acquisition unit is used for acquiring a facial information image according to the position information of the target to be detected;
the mesoscale environment information acquisition unit is used for acquiring an image containing an environment background by using an adaptive method independent of the resolution of a camera and the size of the image according to the position information of the target to be detected;
the high-scale behavior information acquisition unit is used for acquiring an image containing the behavior information of the target to be detected by using an adaptive method independent of the resolution of the camera and the size of the image according to the position information of the target to be detected.
5. The deep learning based multi-scale living detection system of claim 3, wherein the convolutional neural network creation module includes:
the RGB image characteristic information extraction network is used for constructing a multi-level deep neural network and extracting multi-level semantic characteristic information of the object to be detected under different scales;
the RGB image characteristic information classification network is used for constructing a multi-level semantic characteristic information fusion network, fusing the extracted semantic information of the object to be detected and outputting a living body score, wherein the score is between 0 and 1, if the object is a living body, the network output result is 1, and if the object is a non-living body, the network output result is 0.
6. The deep learning based multi-scale biopsy system of claim 3, wherein the multi-scale biopsy module comprises:
a low-scale biological imaging feature constraint unit for giving constraint weight to the facial imaging features extracted from the low-scale image;
the mesoscale environment information feature constraint unit is used for giving constraint weight to the environment features extracted from the mesoscale image;
the high-scale behavior information feature constraint unit is used for giving constraint weights to the behavior features of the target to be detected extracted from the high-scale image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011047776.0A CN112183357B (en) | 2020-09-29 | 2020-09-29 | Multi-scale living body detection method and system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011047776.0A CN112183357B (en) | 2020-09-29 | 2020-09-29 | Multi-scale living body detection method and system based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112183357A CN112183357A (en) | 2021-01-05 |
| CN112183357B true CN112183357B (en) | 2024-03-26 |
Family
ID=73946353
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011047776.0A Active CN112183357B (en) | 2020-09-29 | 2020-09-29 | Multi-scale living body detection method and system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112183357B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114155591A (en) * | 2021-12-02 | 2022-03-08 | 海南伍尔索普电子商务有限公司 | Human face silence living body detection method based on multi-scale information fusion |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109670430A (en) * | 2018-12-11 | 2019-04-23 | 浙江大学 | A kind of face vivo identification method of the multiple Classifiers Combination based on deep learning |
| CN110569808A (en) * | 2019-09-11 | 2019-12-13 | 腾讯科技（深圳）有限公司 | Living body detection method and device and computer equipment |
| CN111144277A (en) * | 2019-12-25 | 2020-05-12 | 东南大学 | Face verification method and system with living body detection function |
| WO2020125623A1 (en) * | 2018-12-20 | 2020-06-25 | 上海瑾盛通信科技有限公司 | Method and device for live body detection, storage medium, and electronic device |
| CN111401134A (en) * | 2020-02-19 | 2020-07-10 | 北京三快在线科技有限公司 | Living body detection method, living body detection device, electronic apparatus, and storage medium |
| CN111696080A (en) * | 2020-05-18 | 2020-09-22 | 江苏科技大学 | Face fraud detection method, system and storage medium based on static texture |
-
2020
- 2020-09-29 CN CN202011047776.0A patent/CN112183357B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109670430A (en) * | 2018-12-11 | 2019-04-23 | 浙江大学 | A kind of face vivo identification method of the multiple Classifiers Combination based on deep learning |
| WO2020125623A1 (en) * | 2018-12-20 | 2020-06-25 | 上海瑾盛通信科技有限公司 | Method and device for live body detection, storage medium, and electronic device |
| CN110569808A (en) * | 2019-09-11 | 2019-12-13 | 腾讯科技（深圳）有限公司 | Living body detection method and device and computer equipment |
| CN111144277A (en) * | 2019-12-25 | 2020-05-12 | 东南大学 | Face verification method and system with living body detection function |
| CN111401134A (en) * | 2020-02-19 | 2020-07-10 | 北京三快在线科技有限公司 | Living body detection method, living body detection device, electronic apparatus, and storage medium |
| CN111696080A (en) * | 2020-05-18 | 2020-09-22 | 江苏科技大学 | Face fraud detection method, system and storage medium based on static texture |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112183357A (en) | 2021-01-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102596897B1 (en) | Method of motion vector and feature vector based fake face detection and apparatus for the same | |
| JP4743823B2 (en) | Image processing apparatus, imaging apparatus, and image processing method | |
| CN111783748B (en) | Face recognition method and device, electronic equipment and storage medium | |
| WO2020195732A1 (en) | Image processing device, image processing method, and recording medium in which program is stored | |
| CN109800643A (en) | A kind of personal identification method of living body faces multi-angle | |
| CN111783629B (en) | Human face in-vivo detection method and device for resisting sample attack | |
| CN111222380B (en) | Living body detection method and device and recognition model training method thereof | |
| JP2006107288A (en) | Personal authentication method, device and program | |
| CN113269010B (en) | Training method and related device for human face living body detection model | |
| CN111767879A (en) | Living body detection method | |
| CN109145704B (en) | Face portrait recognition method based on face attributes | |
| CN112633217A (en) | Human face recognition living body detection method for calculating sight direction based on three-dimensional eyeball model | |
| CN112183357B (en) | Multi-scale living body detection method and system based on deep learning | |
| Kim et al. | Lip print recognition for security systems by multi-resolution architecture | |
| Huang et al. | Multi-Teacher Single-Student Visual Transformer with Multi-Level Attention for Face Spoofing Detection. | |
| Kim et al. | Deep residual network-based recognition of finger wrinkles using smartphone camera | |
| Jindal et al. | Sign Language Detection using Convolutional Neural Network (CNN) | |
| CN113723215B (en) | Training method of living body detection network, living body detection method and device | |
| CN110991223B (en) | Pupil identification method and system based on transfer learning | |
| CN112883831A (en) | Living body detection method and device, electronic equipment and storage medium | |
| Senarathne et al. | Blindaid-android-based mobile application guide for visually challenged people | |
| Lin et al. | 3D face authentication by mutual coupled 3D and 2D feature extraction | |
| CN111126283A (en) | Rapid in-vivo detection method and system for automatically filtering fuzzy human face | |
| CN117351579B (en) | Iris living body detection method and device based on multi-source information fusion | |
| KR102318051B1 (en) | Method for examining liveness employing image of face region including margin in system of user identifying |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |