CN111967296B - Iris living body detection method, access control method and device - Google Patents
Iris living body detection method, access control method and device Download PDFInfo
- Publication number
- CN111967296B CN111967296B CN202010599660.1A CN202010599660A CN111967296B CN 111967296 B CN111967296 B CN 111967296B CN 202010599660 A CN202010599660 A CN 202010599660A CN 111967296 B CN111967296 B CN 111967296B
- Authority
- CN
- China
- Prior art keywords
- face
- iris
- forgery
- detection result
- human
- 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 441
- 238000000034 method Methods 0.000 title claims abstract description 64
- 210000001508 eye Anatomy 0.000 claims abstract description 169
- 230000005855 radiation Effects 0.000 claims abstract description 110
- 238000012549 training Methods 0.000 claims abstract description 109
- 238000013528 artificial neural network Methods 0.000 claims abstract description 58
- 230000036760 body temperature Effects 0.000 claims description 120
- 238000001574 biopsy Methods 0.000 claims description 23
- 210000005252 bulbus oculi Anatomy 0.000 claims description 16
- 238000009826 distribution Methods 0.000 claims description 15
- 238000001727 in vivo Methods 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 13
- 238000000605 extraction Methods 0.000 claims description 12
- 210000001747 pupil Anatomy 0.000 claims description 10
- 238000003860 storage Methods 0.000 claims description 9
- 230000001815 facial effect Effects 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 4
- 210000000554 iris Anatomy 0.000 description 279
- 238000010586 diagram Methods 0.000 description 17
- 238000009529 body temperature measurement Methods 0.000 description 16
- 230000004927 fusion Effects 0.000 description 12
- 230000005764 inhibitory process Effects 0.000 description 12
- 239000011521 glass Substances 0.000 description 10
- 230000008569 process Effects 0.000 description 10
- 230000006870 function Effects 0.000 description 7
- 230000001276 controlling effect Effects 0.000 description 5
- 238000010606 normalization Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 5
- 208000006550 Mydriasis Diseases 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000004861 thermometry Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 210000000887 face Anatomy 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 230000002265 prevention Effects 0.000 description 2
- 240000000015 Iris germanica Species 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000005856 abnormality Effects 0.000 description 1
- 230000003796 beauty Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 210000000744 eyelid Anatomy 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 230000001105 regulatory effect Effects 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
- G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
-
- 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
- G06V10/10—Image acquisition
- G06V10/12—Details of acquisition arrangements; Constructional details thereof
- G06V10/14—Optical characteristics of the device performing the acquisition or on the illumination arrangements
- G06V10/143—Sensing or illuminating at different wavelengths
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/18—Eye characteristics, e.g. of the iris
- G06V40/197—Matching; Classification
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07C—TIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE
- G07C9/00—Individual registration on entry or exit
- G07C9/30—Individual registration on entry or exit not involving the use of a pass
- G07C9/32—Individual registration on entry or exit not involving the use of a pass in combination with an identity check
- G07C9/37—Individual registration on entry or exit not involving the use of a pass in combination with an identity check using biometric data, e.g. fingerprints, iris scans or voice recognition
Abstract
The invention provides an iris living body detection method, an access control method and an access control device, wherein the detection method comprises the following steps: acquiring an image of heat radiation, visible light and near infrared which can be aligned; inputting the multi-mode combination of the visible light image and/or the near infrared image and the thermal radiation image into a multi-mode human face living body detection model to obtain a human face forgery detection result; the multi-mode human face living body detection model is obtained by training a corresponding multi-input neural network through samples of a true human face and a human face pseudo-object type which are combined in a multi-mode manner; performing living body detection on the near infrared image by using a human eye region living body detection model to obtain a human eye region forgery detection result; the human eye region living body detection model is obtained by training a neural network by using samples of human eyes and human eye region pseudo-object types of human eye near infrared image modes; and obtaining iris living body detection results. Through the scheme, the access control management of high-precision iris recognition under the condition of complex face shielding can be realized.
Description
Technical Field
The invention relates to the technical field of identity recognition, in particular to an iris living body detection method, an access control method and an access control device.
Background
At present, the access control is mainly realized by adopting fingerprint identification and face recognition technologies, and the two biological characteristic technologies have lower cost and are convenient to use under general conditions. However, in special application scenes such as hospitals and special periods such as epidemic prevention, people need to wear prevention and control articles such as masks, goggles and hats, so that faces of people can be shielded to different degrees, and in this case, the using convenience of face recognition is seriously reduced, and even the face recognition cannot be correctly recognized at all; and fingerprinting becomes unusable due to contact limitations. The iris recognition has no contact, and shielding objects such as masks, glasses, goggles, hats and the like do not need to be removed, so the iris recognition is more suitable for the identity recognition under the special condition.
Disclosure of Invention
The invention provides an iris living body detection method, an access control method and an access control device, which are used for realizing access control management of high-precision iris recognition under the condition of complex face shielding.
In order to achieve the above purpose, the invention is realized by adopting the following scheme:
according to an aspect of an embodiment of the present invention, there is provided an iris in vivo detection method including:
Acquiring a thermal radiation image, a visible light image and a near infrared image of an object to be detected; the thermal radiation image, the visible light image and the face region in the near infrared image can be aligned;
inputting a multi-mode combined face area image obtained according to at least one of the visible light image and the near infrared image and the thermal radiation image into a multi-mode face living body detection model to obtain a multi-mode face forgery detection result; the multi-mode human face living body detection model is obtained by training a first neural network with corresponding multiple inputs through a first training sample conforming to the multi-mode combination, and the set of the first training samples comprises training samples of a true human face type and a human face pseudo-object type;
performing living body detection on a human eye region image in the near infrared image by using a human eye region living body detection model to obtain a human eye region forgery detection result; the human eye region living body detection model is obtained by training a second neural network by using a second training sample conforming to a human eye near infrared image mode, and the set of the second training samples comprises training samples of a true human eye type and a human eye region forgery type; the human eye region artifact type includes an iris artifact type;
And obtaining an iris living body detection result of the object to be detected according to the multi-mode face forgery detection result and the human eye area forgery detection result.
In some embodiments, in a case of a multi-modal combined face area image obtained from at least one of the visible light image and the near infrared image and the thermal radiation image, the first neural network is a multi-channel input network; or, in the case of a multi-mode combined face area image obtained by combining one of the visible light image and the near infrared image with the thermal radiation image, the first neural network is a twin network.
In some embodiments, acquiring a thermal radiation image, a visible light image, and a near infrared image of an object to be detected includes: collecting a visible light image comprising the object to be detected, and detecting the position of a face area in the visible light image to obtain the face area image in the visible light image; according to the position of a human face region in the visible light image, controlling and collecting a near infrared image comprising a clear iris of the object to be detected, obtaining a human face region image and a human eye region image in the near infrared image, and carrying out iris positioning on the human eye region image in the near infrared image to obtain an iris region image in the near infrared image; and acquiring a thermal radiation image comprising the object to be detected, and mapping the face region image in the visible light image or the near infrared image to the thermal radiation image according to the position relation of the related camera to obtain the face region image in the thermal radiation image.
In some embodiments, the eye region artifact type further comprises a face artifact type; the second neural network comprises a first convolution network layer, a second convolution network layer, a first output layer and a second output layer; the output end of the first convolution network layer is connected with the first output layer and is used for extracting characteristics of a human eye region image in the received near infrared image and outputting human eye region forgery detection results of human face forgery types according to the characteristic extraction results; the input end and the output end of the second convolution network layer are respectively connected with the output end and the second output layer of the first convolution network layer, and are used for carrying out feature extraction on iris region features obtained from feature extraction results of human eye region images in the near infrared images according to iris region positioning results in the human eye region images in the near infrared images, and outputting human eye region forgery detection results of iris forgery types according to feature extraction results of the iris region features.
In some embodiments, the iris biopsy method further comprises: performing living body detection on the thermal radiation image by using a human body temperature living body detection model to obtain a human face forgery detection result based on the human body temperature; the human body temperature living body detection model is obtained by training a third neural network by using a third training sample conforming to a thermal radiation image mode, and the set of the third training sample comprises training samples of a real human face type and a human face pseudo-object type. Obtaining an iris living body detection result of the object to be detected according to the multi-mode face forgery detection result and the human eye area forgery detection result, wherein the iris living body detection result comprises the following steps: and obtaining an iris living body detection result of the object to be detected according to the multi-mode human face false body detection result, the human eye region false body detection result and the human face false body detection result based on the human body temperature.
In some embodiments, the iris biopsy method further comprises: inputting a multi-mode combined iris region image obtained according to at least one of the visible light image and the near infrared image and the thermal radiation image into a multi-mode iris living body detection model to obtain a multi-mode iris forgery detection result; the multi-modal iris living detection model is obtained by training a fourth neural network with corresponding multi-inputs by using a fourth training sample conforming to the multi-modal combination, wherein the set of the fourth training samples comprises training samples of a true human iris type and an iris pseudo-object type, and the fourth neural network is identical to the first neural network. Obtaining an iris living body detection result of the object to be detected according to the multi-mode face forgery detection result and the human eye area forgery detection result, wherein the iris living body detection result comprises the following steps: and obtaining an iris living body detection result of the object to be detected according to the multi-mode face forgery detection result, the human eye area forgery detection result and the multi-mode iris forgery detection result.
In some embodiments, the iris biopsy method further comprises: performing living body detection on the thermal radiation image by using a human body temperature living body detection model to obtain a human face forgery detection result based on the human body temperature; the human body temperature living body detection model is obtained by training a third neural network by using a third training sample conforming to a thermal radiation image mode, and the set of the third training sample comprises training samples of a real human face type and a human face pseudo-object type; inputting a multi-mode combined iris region image obtained according to at least one of the visible light image and the near infrared image and the thermal radiation image into a multi-mode iris living body detection model to obtain a multi-mode iris forgery detection result; the multi-modal iris living detection model is obtained by training a fourth neural network with corresponding multi-inputs by using a fourth training sample conforming to the multi-modal combination, wherein the set of the fourth training samples comprises training samples of a true human iris type and an iris pseudo-object type, and the fourth neural network is identical to the first neural network. Obtaining an iris living body detection result of the object to be detected according to the multi-mode face forgery detection result and the human eye area forgery detection result, wherein the iris living body detection result comprises the following steps: normalizing and weighting and summing the human face false object detection result based on the human body temperature, the multi-mode human face false object detection result and the human eye area false object detection result of the human face false object type to obtain a detection result that the object to be detected is the human face false object; wherein the facial artifact comprises one or more of an unmanned body temperature prosthesis, an artificial heat source body, and a worn mask, the unmanned body temperature prosthesis comprising one or more of a printed paper and a mannequin; normalizing and weighting summing the multi-mode iris artifact detection result and the human eye region artifact detection result of the iris artifact type to obtain a detection result that the object to be detected is an iris artifact; iris artifacts include one or more of a pupil, an artificial eye, and a vitreous eyeball to be worn; and outputting a detection result that the object to be detected is a human face artifact and/or a detection result that the object to be detected is an iris artifact as an iris living detection result of the object to be detected.
In some embodiments, the iris biopsy method further comprises: performing living body detection on the thermal radiation image by using a human body temperature living body detection model to obtain a human face forgery detection result based on the human body temperature; the human body temperature living body detection model is obtained by training a third neural network by using a third training sample conforming to a thermal radiation image mode, and the set of the third training sample comprises training samples of a real human face type and a human face pseudo-object type; inputting a multi-mode combined iris region image obtained according to at least one of the visible light image and the near infrared image and the thermal radiation image into a multi-mode iris living body detection model to obtain a multi-mode iris forgery detection result; the multi-modal iris living detection model is obtained by training a fourth neural network with corresponding multi-inputs by using a fourth training sample conforming to the multi-modal combination, wherein the set of the fourth training samples comprises training samples of a true human iris type and an iris pseudo-object type, and the fourth neural network is identical to the first neural network. Obtaining an iris living body detection result of the object to be detected according to the multi-mode face forgery detection result and the human eye area forgery detection result, wherein the iris living body detection result comprises the following steps: if the normalized score of the human face false artifact detection result based on the human body temperature is greater than or equal to a first set human face false artifact score threshold value or the normalized score of the human eye area false artifact detection result of the human face false artifact type is greater than or equal to a second set human face false artifact score threshold value, determining that the object to be detected is a human face false artifact; if the normalized score of the human face false artifact detection result based on the human body temperature is smaller than the first set human face false artifact score threshold and larger than or equal to a third set human face false artifact score threshold, and the normalized score of the human eye region false artifact detection result of the human face false artifact type is smaller than the second set threshold and larger than or equal to a fourth set threshold, determining that the object to be detected is a human face false artifact; wherein the third set threshold is less than the first set threshold and the fourth set threshold is less than the second set threshold;
If the normalized score of the human body temperature-based face forgery detection result is smaller than the first set face forgery score threshold and larger than or equal to the third set face forgery score threshold, and the normalized score of the multi-mode face forgery detection result is smaller than the fifth set face forgery score threshold and larger than or equal to the sixth set face forgery score threshold, determining that the object to be detected is a face forgery; wherein the fifth set face forgery score threshold is less than the first set face forgery score threshold and the second set face forgery score threshold, and the sixth set face forgery score threshold is less than the fifth set face forgery score threshold;
if the normalized score of the human body temperature-based human face forgery detection result is smaller than the third set human face forgery score threshold and larger than or equal to a seventh set human face forgery score threshold, and the normalized score of the human eye region forgery detection result of the human face forgery type is smaller than the fourth set human face forgery score threshold and larger than or equal to an eighth set human face forgery score threshold, and the normalized score of the multi-mode human face forgery detection result is smaller than the sixth set human face forgery score threshold and larger than or equal to a ninth set human face forgery score threshold, determining that the object to be detected is human face forgery; the seventh set face forgery score threshold is smaller than the third set face forgery score threshold, the eighth set face forgery score threshold is smaller than the fourth set face forgery score threshold, and the ninth set face forgery score threshold is smaller than the sixth set face forgery score threshold;
If the normalized score of the human eye region forgery detection result of the iris forgery type is larger than or equal to a first set iris forgery score threshold value, determining that the object to be detected is iris forgery;
if the normalized score of the human eye region false object detection result of the iris false object type is smaller than the first set iris false object score threshold and larger than or equal to the second set iris false object score threshold, and the normalized score of the multi-mode iris false object detection result is smaller than the third set iris false object score threshold and larger than or equal to the fourth set iris false object score threshold, determining that the object to be detected is iris false object; the second set iris artifact score threshold is less than the first set iris artifact score threshold, the fourth set iris artifact score threshold is less than the third set iris artifact score threshold, and the third set iris artifact score threshold is less than the first set iris artifact score threshold.
In some embodiments, the third set face artifact score threshold is a first set fractional multiple of the first set face artifact score threshold, the fourth set face artifact score threshold is a first set fractional multiple of the second set face artifact score threshold, and the sixth set face artifact score threshold is a first set fractional multiple of the fifth set face artifact score threshold; the seventh set face forgery score threshold is a second set fractional multiple of the first set face forgery score threshold, the eighth set face forgery score threshold is a second set fractional multiple of the second set face forgery score threshold, and the ninth set face forgery score threshold is a second set fractional multiple of the fifth set face forgery score threshold; the second set fractional multiple is smaller than the first set fractional multiple; the second set iris artifact score threshold is a first set fractional multiple of the first set iris artifact score threshold and the fourth set iris artifact score threshold is a first set fractional multiple of the third set iris artifact score threshold.
In some embodiments, the iris biopsy method further comprises: acquiring the body temperature of the object to be detected, and outputting body temperature alarm information or iris living body detection result that the object to be detected is forgery under the condition that the body temperature of the object to be detected exceeds a set human body temperature threshold range.
In some embodiments, obtaining the body temperature of the subject to be tested comprises: performing face recognition on the object to be detected by utilizing the visible light image and/or the near infrared image to obtain a face key point positioning result; determining the face shielding type of the object to be detected according to the positioning result of the face key points; the face shielding type is that the face is not shielded, the mask is worn, the goggles are worn, and the mask and the goggles are worn; if the face shielding type of the object to be detected is that the face is not shielded, calculating the average temperature of the whole face area of the object to be detected according to the temperature distribution data of the object to be detected, and taking the average temperature as the body temperature of the object to be detected; wherein the temperature distribution data of the object to be detected corresponds to the thermal radiation image; if the face shielding type of the object to be detected is shielding of a wearing mask, calculating the average temperature of a face area above the mask worn by the object to be detected according to the temperature distribution data of the object to be detected, and taking the average temperature as the body temperature of the object to be detected; if the face shielding type of the object to be detected is shielding of wearing goggles, calculating the average temperature of a face area below the goggles worn by the object to be detected according to the temperature distribution data of the object to be detected, and taking the average temperature as the body temperature of the object to be detected; and if the face shielding type of the object to be detected is shielding comprising a wearing mask and goggles, obtaining the temperature of the exposed skin at the upper edge of the goggles worn by the object to be detected according to the temperature distribution data of the object to be detected, and taking the temperature as the body temperature of the object to be detected.
In some embodiments, inputting a multi-modal combined face area image obtained from at least one of the visible light image and the near infrared image and the thermal radiation image into a multi-modal face living body detection model to obtain a multi-modal face artifact detection result, including: and under the condition that the body temperature of the object to be detected does not exceed the set human body temperature threshold range, inputting a multi-mode combined face area image obtained according to at least one of the visible light image and the near infrared image and the thermal radiation image into a multi-mode face living body detection model to obtain a multi-mode face forgery detection result. Performing living body detection on a human eye region image in the near infrared image by using a human eye region living body detection model to obtain a human eye region forgery detection result, wherein the method comprises the following steps: and under the condition that the body temperature of the object to be detected does not exceed the set human body temperature threshold range, performing living detection on a human eye region image in the near infrared image by using a human eye region living detection model to obtain a human eye region forgery detection result.
According to another aspect of the embodiment of the invention, there is provided an iris recognition entrance guard control method, including: obtaining an iris living body detection result of the object to be detected by using the iris living body detection method of any embodiment; and controlling an entrance guard switch according to the iris living body detection result of the object to be detected.
According to another aspect of the embodiment of the present invention, there is provided an iris biopsy apparatus including a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the iris biopsy method according to any of the above embodiments when executing the program.
According to another aspect of the embodiment of the present invention, there is provided an iris recognition entrance guard control system, including: the visible light camera is used for collecting visible light images of the object to be detected; the near infrared camera is used for collecting near infrared images of the object to be detected; the temperature measuring module is used for collecting a thermal radiation image of an object to be detected; the iris living body detection device according to any of the above embodiments is configured to obtain an iris living body detection result of an object to be detected according to a visible light image, a near infrared image, and a thermal radiation image of the object to be detected.
According to another aspect of the embodiment of the present invention, there is provided an iris recognition access control system, including: the iris recognition entrance guard control system according to any one of the above embodiments.
According to another aspect of an embodiment of the present invention, there is provided a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the method of any of the embodiments described above.
The iris living detection method, the iris recognition gate inhibition control method, the iris living detection device, the iris recognition gate inhibition control system, the iris recognition gate inhibition system and the computer readable storage medium can realize gate inhibition control management of high-precision iris recognition under the condition of complex face shielding by realizing high-reliability living detection of multi-mode fusion by adopting thermal radiation images.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art. In the drawings:
FIG. 1 is a flow chart of an iris in vivo detection method according to an embodiment of the invention;
FIG. 2 is a diagram illustrating a network structure of a human eye region living body detection model according to an embodiment of the present invention;
FIG. 3 is a flow chart of an iris recognition entrance guard control method according to an embodiment of the invention;
Fig. 4 is a schematic structural diagram of an iris recognition entrance guard control system according to an embodiment of the invention;
FIG. 5 is a flow chart of an iris recognition entrance guard control method according to an embodiment of the invention;
FIG. 6 is a schematic diagram of a twin network in an embodiment of the present invention;
FIG. 7 is a schematic diagram of a multi-channel input network in accordance with an embodiment of the invention;
fig. 8 is a network structure diagram of a human eye region living body detection model according to an embodiment of the invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the embodiments of the present invention will be described in further detail with reference to the accompanying drawings. The exemplary embodiments of the present invention and their descriptions herein are for the purpose of explaining the present invention, but are not to be construed as limiting the invention.
In large scale epidemic outbreaks, body temperature is often one of the most direct physical health indicators, and is also one of the important means to ensure that the identified target is from a living human. Identification, body temperature detection and high-reliability living body detection play an important role in regional control.
In this regard, the embodiment of the invention provides an iris living body detection method, which creatively combines a thermal radiation image capable of reflecting the body temperature of a human body with other types of images to carry out living body detection, so that access control management of high-precision iris recognition under the condition that the face is blocked in an epidemic period can be realized.
Fig. 1 is a flow chart of an iris biopsy method according to an embodiment of the present invention. Referring to fig. 1, the iris in vivo detection method of the embodiments may include the following steps S110 to S140. The execution sequence of step S120 and step S130 may be interchanged.
It should be noted in advance that a face artifact may refer to a prosthesis of a larger area (e.g., greater than or equal to the face), and an iris artifact may refer to a prosthesis within the vicinity of the iris. The coverage of the face artifact is often larger than that of the iris artifact, and the prosthesis types of the face artifact and the iris artifact are different due to the fact that the coverage of the face artifact and the iris artifact are different. For example, the face artifact may be a printing paper, a face model, a mask, or the like, and the iris artifact may be a pupil, an artificial eye, a glass eyeball, or the like. In addition, the human eye region artifact may include an iris artifact; the human eye region artifact may also include an eye artifact of an area including the periocular region and the eyeball, and since the iris or the eyeball region is exceeded and also includes the periocular portion, the type of the human eye artifact tends to be similar to or identical with that of a human face artifact, so it can be said that the human eye region artifact may also include the type of a human face artifact.
Specific embodiments of step S110 to step S140 will be described in detail below.
Step S110: acquiring a thermal radiation image, a visible light image and a near infrared image of an object to be detected; wherein the thermal radiation image, the visible light image, and the face region in the near infrared image can be aligned.
In the above step S110, the object to be detected may refer to a human body part including a human face region. The thermal radiation image can be acquired by a temperature measuring module, the visible light image can be acquired by a visible light camera (a face camera), and the near infrared image can be acquired by a near infrared camera. In addition, the iris recognition device can comprise a visible light camera and a near infrared image, and a temperature measurement module can be added in the iris recognition device to collect a thermal radiation image, a visible light image and a near infrared image. In addition, the object to be detected by living body detection can continue iris identification by using the iris identification device in the iris identification device.
The fact that the thermal radiation image, the visible light image and the face region in the near infrared image can be aligned may mean that the positions of the face regions in the various images can be directly or after deformation, which requires that the relative positions of the devices for acquiring the corresponding images are relatively consistent or fixed. For example, the positions of the visible light camera and the temperature measuring module are relatively fixed, and the images acquired by the two are relatively easy to align, so in the subsequent step S120, in order to conveniently use the existing equipment, a multi-mode combination can be formed by using the visible light image and the thermal radiation image.
In the implementation, in the step S110, a thermal radiation image, a visible light image, and a near infrared image of the object to be detected are obtained, which may specifically include the steps of: s111, collecting a visible light image comprising the object to be detected, and detecting the position of a face area in the visible light image to obtain the face area image in the visible light image; s112, controlling and collecting near infrared images including clear irises of the objects to be detected according to the positions of face areas in the visible light images to obtain face area images and eye area images in the near infrared images, and performing iris positioning on the eye area images in the near infrared images to obtain iris area images in the near infrared images; s113, acquiring a thermal radiation image comprising the object to be detected, and mapping the face area image in the visible light image or the near infrared image to the thermal radiation image according to the position relation of the related camera to obtain the face area image in the thermal radiation image.
In the step S111, a visible light image may be acquired by using a visible light camera, and then face detection is performed to obtain a face image in the visible light image; or the visible light face camera can be used for directly acquiring the visible light face image. In the step S112, according to the detected face position in the visible light image, the acquisition posture of the near-infrared camera may be adjusted to enable the iris to be clear, so as to obtain a near-infrared image, specifically, a near-infrared face image and a near-infrared eye area image, and further iris positioning may be performed, so as to obtain an iris area image in the visible light image. In the step S113, after the thermal radiation image of the object to be detected is acquired, the face area image in the thermal radiation image may be obtained by mapping, so as to obtain the face temperature.
In addition, in the step S111, the location of the key points of the visible light image may be further performed to obtain the positions of the key points (such as eyes and nose) of the face. In the step S112, the near infrared image may be further subjected to the key point positioning to obtain the position of the key point (e.g. the corner of the eye) of the human eye. The face key points and the eye key points can be used for determining the face shielding type of the object to be detected, such as whether shielding exists, whether a mask is worn, and whether goggles are worn.
In this embodiment, the existing method may be used to identify the face position and the face key point from the visible light image and the near infrared image, and the similar method may also be used to identify the eye region position and the eye key point from the near infrared image. The position of the face area is difficult to find by independently utilizing the thermal radiation image, and the acquisition position can be aligned by controlling the acquisition process, so that the face area in the acquired image can be aligned, the face position can be mapped to the thermal radiation image, and the accurate face position in the thermal radiation image is obtained.
Step S120: inputting a multi-mode combined face area image obtained according to at least one of the visible light image and the near infrared image and the thermal radiation image into a multi-mode face living body detection model to obtain a multi-mode face forgery detection result; the multi-mode human face living body detection model is obtained by training a first neural network with corresponding multiple inputs through a first training sample conforming to the multi-mode combination, and the set of the first training samples comprises training samples of a true human face type and a human face pseudo-object type.
In step S120, the multi-modal combination may refer to a combination of multiple information. Each of the visible light image, the near infrared image, and the thermal radiation image contains one image information, each of which may be referred to as one modality. The multi-modal combination may include one of a visible light image and a near infrared image, or both of a visible light image and a near infrared image, and further includes a thermal radiation image. If the visible light image, the near infrared image and the thermal radiation image are all face images, the multi-mode combined face area image can be directly obtained by combining the images; if the images include other areas in addition to the face area, the face area images of the respective images can be obtained through the previous or subsequent image recognition, and then the multi-mode combined face area images are formed. The multi-modal combination is obtained from three images, i.e., a visible light image, a near infrared image, and a thermal radiation image, and does not exclude that other types of images may be included.
The multi-mode human face living body detection model can receive input of information of multiple modes, or can take the information of the multiple modes into consideration, and the detection result is obtained through analysis. The first neural network on which the multi-modal face living detection model is based may be a convolutional neural network, and may be a combination of multiple networks. The output result of the multi-mode face living body detection model may be a classification result, may be a probability or a score, and is, for example, a probability of face forgery. In addition, the recognition objects or capabilities of the model may vary from training sample to training sample. The first training sample conforming to the multi-modality combination may mean that the first training sample contains various information in the multi-modality combination, for example, if the multi-modality combination includes a thermal radiation image modality and a visible light image modality, one first training sample may contain a thermal radiation image and a visible light image of the same object (e.g., a face or a prosthesis) and a corresponding label whether or not it is a face artifact. The first training sample set may include a plurality of training samples, for example, a sample of a true face type and a sample of a face forgery type. The training samples of the human face forgery type may include, for example, training samples of the type of a human body temperature-free prosthesis, an artificial heat source body, a mask to be worn, or the like, wherein the human body temperature-free prosthesis may be, for example, a printed paper of a human face, may be a human body model of an unmanned body temperature, and the artificial heat source body may be a human body model with a heat source. In addition, the "first" in the first training sample is to distinguish from the samples used to train other neural networks, and does not refer to the order of the samples in the same training sample set, etc.
In the step S120, the first neural network may be various multi-input neural networks.
For example, in the case of a multi-modal combined face area image obtained from at least one of the visible light image and the near infrared image and the thermal radiation image, the first neural network may be a multi-channel input network. The multi-channel input network may also be called pseudo-twin network, and may include multiple parts of the network according to the number of channels, each part of the network (may include a convolution layer, a pooling layer, and a full connection layer) may separately receive corresponding inputs and process the inputs, and output data of the full connection layers of each part of the network may be fused together to form a result considering the multi-channel inputs. When the model is trained, a loss function can be obtained after a result considering the multi-channel input is obtained, and then the loss function is fed back to the model, so that the aim of training the multi-channel input network is fulfilled. If the multi-modal combination consists of a visible light image (or near infrared image) and a thermal radiation image, the multi-channel input network may have two-channel inputs; if the multi-modal combination consists of a visible light image, a near infrared image and a thermal radiation image, the multi-channel input network may have three channels of inputs.
For another example, in a case of a face region image obtained by multi-modal combination of one of the visible light image and the near infrared image and the thermal radiation image, the first neural network may be a twin network. The main difference between the twin network and the multi-channel input network (pseudo-twin network) described above is that there is a certain correlation of channel parameters of the pooling layer of the two-part network in the twin network.
Step S130: performing living body detection on a human eye region image in the near infrared image by using a human eye region living body detection model to obtain a human eye region forgery detection result; the human eye region living body detection model is obtained by training a second neural network by using a second training sample conforming to a human eye near infrared image mode, and the set of the second training samples comprises training samples of a true human eye type and a human eye region forgery type; the human eye region artifact type includes an iris artifact type.
In the above step S130, the human eye region artifact type includes an iris artifact type, so that the second neural network may be trained using the second training sample of the type to identify whether the object to be detected belongs to an iris artifact, so that it may be detected whether to wear a prosthesis attack of the iris artifact type such as pupil, eye prosthesis, glass eyeball, etc.
As such, the categories of training samples included in the second set of training samples may include true human eye types and various human eye region artifact types. The type of face artifact may be, for example, a prosthesis without a human body temperature, an artificial heat source body, a mask to be worn, or the like, and for the type of human eye region artifact, the type of face artifact mainly refers to a human eye region (may include eyelid around eyeball, canthus, or the like). The human eye type may refer to information about human eye areas of a human being, a non-artifact.
Further, the human eye region artifact type may include both a face artifact type and an iris artifact type, whereby it is possible to detect whether it is an iris artifact or not using not only the iris region image but also the human eye region image.
Fig. 2 is a network structure diagram of a human eye region living body detection model according to an embodiment of the invention, referring to fig. 2, the second neural network may include a first convolutional network layer, a second convolutional network layer, a first output layer, and a second output layer. The output end of the first convolution network layer is connected with the first output layer and can be used for extracting characteristics of a human eye region image in the received near infrared image and outputting human eye region forgery detection results of human face forgery types according to the characteristic extraction results. The input end and the output end of the second convolution network layer are respectively connected with the output end and the second output layer of the first convolution network layer, and can be used for extracting features of iris areas obtained from feature extraction results of human eye area images in the near infrared images according to iris area positioning results in the human eye area images in the near infrared images, and outputting human eye area forgery detection results of iris forgery types according to the feature extraction results of the iris area features.
In this embodiment, the iris region positioning result in the human eye region image in the near infrared image may be obtained by additionally performing iris positioning on the near infrared image, so that the second convolution network layer may also receive an iris positioning result input from the outside. Through the connection relation between the first convolution network layer and the second convolution network layer, the second neural network can share the features, on one hand, the features extracted from the human eye region image can be all used for identifying the human face artifacts, and on the other hand, the features extracted from the iris region in the human eye region image can be used for identifying the iris artifacts.
Step S140: and obtaining an iris living body detection result of the object to be detected according to the multi-mode face forgery detection result and the human eye area forgery detection result.
In the above step S140, the human eye region artifact may include an iris artifact type, or may further include a face artifact type. The multi-mode human face forgery detection result and human eye area forgery detection result of the human face forgery type can be used for judging whether an object to be detected is human face forgery (the attack range of a prosthesis is large, such as a mask, printing paper and the like); the human eye region forgery detection result of the iris forgery type can be used to determine whether or not the object to be detected is an iris forgery (the prosthesis attack is mainly in the iris or eyeball range, such as pupil, artificial eye, etc.). In general, if it is determined that the object to be detected is a face artifact or an iris artifact, it can be considered that the living body detection is not passed. In this way, whether the result of the face forgery is a face forgery or not can be obtained by integrating the various face forgery detection results, whether the result of the iris forgery is obtained by integrating the various iris forgery detection results, and whether the result of the living body is obtained by integrating the two types of results.
In the above embodiments, images of various modalities required for the subsequent steps may be obtained through the above step S110; by the step S120, it is possible to detect whether the object to be detected belongs to a false human face type of false human face type by using multi-mode (multi-source) images, and the false human face detection result with higher accuracy can be obtained due to the fusion of the information of at least two modes; through the above step S130, it is possible to detect whether the object to be detected belongs to the false body attack of the iris false body type based on the near infrared image, and further, it is also possible to obtain the detection result of the false body of the face based on the near infrared image through the feature sharing; through the step S140, the multi-mode detection result and the detection result based on the near infrared image can be considered at the same time, so that more information is fused to obtain the final living body detection result of the object to be detected, and the living body detection precision is further improved.
In order to further improve the accuracy of iris living body detection, human face forgery detection can be performed based on the human body temperature, and living body detection results can be obtained by combining various detection results.
In specific implementation, the iris biopsy method shown in fig. 1 may further include the steps of: s150, performing living body detection on the thermal radiation image by using a human body temperature living body detection model to obtain a human face forgery detection result based on the human body temperature; the human body temperature living body detection model is obtained by training a third neural network by using a third training sample conforming to a thermal radiation image mode, and the set of the third training sample comprises training samples of a real human face type and a human face pseudo-object type.
In this case, in some embodiments, the final living body detection result may be obtained by combining the various detection results of step S120, step S130, and step S150. For example, the step S140 may be specifically performed to obtain the iris living body detection result of the object to be detected according to the multi-mode face forgery detection result and the human eye area forgery detection result, where the step may include the steps of: s1411, obtaining an iris living body detection result of the object to be detected according to the multi-mode face false detection result, the human eye region false detection result and the human body temperature-based face false detection result. In the step S1411, the multi-mode face artifact detection result and the face artifact detection result based on the body temperature may be normalized and weighted to obtain a face artifact detection result; taking the human eye region artifact detection result (for the second training sample for the iris artifact type) as an iris artifact detection result; then, if only one of the face forgery detection result and the iris forgery detection result indicates a prosthesis, the final detection result may be considered as a prosthesis attack, not a living body.
In this embodiment, based on the detection results of the steps S120 and S130, the final iris living body detection result is obtained by combining the detection results based on the facial body temperature, so that more modal information is considered, and the detection accuracy is further improved.
In other embodiments, whether the object is a forgery or not may be determined according to the detection result in the step S150, if yes, it may be determined directly that the object to be detected is not a living body, and the subsequent steps may not be performed. Because the false artificial body is difficult to achieve the effect of realistic human body temperature, the body temperature result obtained according to the thermal radiation image is directly used for judging whether the false body is the false body or not, and the judging efficiency can be improved.
In order to further improve the accuracy of the iris in-vivo detection result, in the case of multi-modal combination for multi-modal face forgery detection in which the resolution of the image is high enough and the iris area is clear enough, multi-modal iris forgery detection can be performed by a method similar to the aforementioned step S130, so that more information can be considered in the determination of iris forgery, improving the detection accuracy.
In specific implementation, the iris biopsy method shown in fig. 1 may further include the steps of: s160, inputting a multi-mode combined iris region image obtained according to at least one of the visible light image and the near infrared image and the thermal radiation image into a multi-mode iris living body detection model to obtain a multi-mode iris forgery detection result; the multi-modal iris living detection model is obtained by training a fourth neural network with corresponding multi-inputs by using a fourth training sample conforming to the multi-modal combination, wherein the set of the fourth training samples comprises training samples of a true human iris type and an iris pseudo-object type, and the fourth neural network is identical to the first neural network.
In this case, the step S140 described above, that is, the iris living body detection result of the object to be detected is obtained from the multi-mode face forgery detection result and the human eye area forgery detection result, may specifically include the steps of: s1421, obtaining an iris living body detection result of the object to be detected according to the multi-mode face forgery detection result, the human eye area forgery detection result and the multi-mode iris forgery detection result. In the step S1421, specifically, the human eye region artifact detection result (for the second training sample for the human face artifact) and the multi-mode iris artifact detection result may be normalized and weighted to obtain the human face artifact detection result; then, if only one of the face forgery detection result and the multi-mode iris forgery detection result indicates a prosthesis, the final detection result may be considered as a prosthesis attack, not a living body.
In the step S160, the neural network on which the multi-mode iris living detection model and the multi-mode face living detection model are based may be the same network, and the main difference is that the fourth training sample used in the multi-mode iris living detection model is for an iris region, the artifact type is an iris artifact type, and the first training sample used in the multi-mode face living detection model is for a face region, and the artifact type is a face artifact type. Therefore, the multi-modal iris living detection model is mainly different from the multi-modal face living detection model in model parameters.
In the step S140, the iris living detection result can be obtained by combining all the existing detection results, and since the multi-mode iris pseudo-object detection result obtained based on the multi-mode iris living detection model is further considered, more information is fused, and the detection accuracy is improved.
For the situation that the final iris living body detection result is obtained based on multiple detection results, the detection results of the type of forgery can be obtained by weighting the detection results of the same type, and further the final living body detection result is obtained by integrating the detection results of different types of forgery. As described above, the steps S1411 and S1421 are specifically implemented.
For another example, in one embodiment, the iris biopsy method shown in fig. 1 may further include: step S150, namely, performing living body detection on the thermal radiation image by using a living body detection model of human body temperature to obtain a human face forgery detection result based on human body temperature; the human body temperature living body detection model is obtained by training a third neural network by using a third training sample conforming to a thermal radiation image mode, and the set of the third training sample comprises training samples of a real human face type and a human face pseudo-object type; the step S160 is also included, namely, an iris area image which is obtained by multi-modal combination according to at least one of the visible light image and the near infrared image and the thermal radiation image is input into a multi-modal iris living body detection model, and a multi-modal iris forgery detection result is obtained; the multi-modal iris living detection model is obtained by training a fourth neural network with corresponding multi-inputs by using a fourth training sample conforming to the multi-modal combination, wherein the set of the fourth training samples comprises training samples of a true human iris type and an iris pseudo-object type, and the fourth neural network is identical to the first neural network. The first neural network may include the first convolutional neural network, the first output layer, the second convolutional neural network, and the second output layer. In this embodiment, the detection results of the multiple modes of steps S120, S130, S150, and S160 are considered at the same time, and the obtained detection results are more accurate.
In this case, in some embodiments, the living body detection result may be obtained by weighting the final result of the detection result. For example, the step S140 may be specifically performed to obtain the iris living body detection result of the object to be detected according to the multi-mode face forgery detection result and the human eye area forgery detection result, where the step may include the steps of:
s1431, detecting the human face forgery based on the human body temperature (expressed as) The multi-modal face artifact detection result (expressed as +.>) And human eye region forgery detection results (expressed as) Carrying out normalization and weighted summation to obtain a detection result of the object to be detected being the face forgery (for example,where α, β, λ may be referred to as weights); wherein the facial artifact may comprise one or more of an unmanned body temperature prosthesis, an artificial heat source body, and a worn mask, the unmanned body temperature prosthesis may comprise one or more of a printed paper and a mannequin;
s1432, detecting the multi-mode iris forgery (expressed as) And iris artifact type human eye region artifact detection result (expressed as +.>) Normalizing and weighting summation is carried out to obtain a detection result (/ -for the iris artifact of the object to be detected >Wherein γ, χ may be referred to as a weight); the iris artifact may include a worn mydriasis,One or more of a prosthetic eye, and a vitreous eyeball;
s1433, the detection result of the object to be detected as the face artifact (expressed as S peri ) And/or the object to be detected is the detection result of iris forgery (which can be expressed as S ocular ) And outputting the iris living body detection result as the iris living body detection result of the object to be detected.
The above-described step S1431 and step S1432 may not limit the execution order. In step S1433, if the face artifact is detected as a result S peri And detection result S of iris forgery ocular One of them reaches the condition of judging as a prosthesis, then the object to be detected can be considered as a prosthesis attack.
In other embodiments, the biopsy result may be obtained by formulating a decision strategy. For example, the step S140 may be specifically performed to obtain the iris living body detection result of the object to be detected according to the multi-mode face forgery detection result and the human eye area forgery detection result, where the step may include the steps of:
s1441, if the normalized score of the human face artifact detection result based on the human body temperature (expressed as) Greater than or equal to a first set face artifact count threshold (expressed as +. >) Or human face artifact type (expressed as +.>) Greater than or equal to a second set face artifact count threshold (expressed as) Determining the object to be detected as a face artifact;
s1442, if the normalized score of the human face artifact detection result based on the human body temperature (expressed as) Less than the first set face artifact count threshold (expressed as +.>) And greater than or equal to a third set face artifact count threshold (which may be expressed as +.>) And the normalized score of the human eye region forgery detection result of the human face forgery type (expressed as +.>) Is smaller than the second set threshold (expressed as +.>) And greater than or equal to a fourth set threshold (which may be expressed as) Determining the object to be detected as a face artifact; wherein the third set threshold is less than the first set threshold and the fourth set threshold is less than the second set threshold;
s1443, if the normalized score of the human face artifact detection result based on the human body temperature (expressed as) Less than the first set face artifact count threshold (expressed as +.>) And greater than or equal to the third set face artifact count threshold (which may be expressed as +. >) And the normalized score of the multi-modal face artifact detection result (expressed as) Less than a fifth set face forgery score threshold (expressed as +.>) And greater than or equal to a sixth set face artifact count threshold (expressed as +.>) Determining the object to be detected as a face artifact; wherein the fifth set face forgery score threshold is less than the first set face forgery score threshold and the second set face forgery score threshold, and the sixth set face forgery score threshold is less than the fifth set face forgery score threshold;
s1444, if the normalized score of the human face artifact detection result based on the human body temperature (expressed as) Less than the third set face artifact count threshold (representable as +.>) And greater than or equal to a seventh set face artifact count threshold (which may be expressed as +.>) And the normalized score of the human eye region forgery detection result of the human face forgery type (expressed as +.>) Less than the fourth set face artifact count threshold (representable as +.>) And greater than or equal to an eighth set face artifact count threshold (which may be expressed as +.>) And the normalized score of the multimode face forgery detection result (expressed as +. >) Less than the sixth set face forgery fraction threshold (expressed as +.>) And greater than or equal to a ninth set face forgery fraction threshold (expressed as +.>) Determining the object to be detected as a face artifact; the seventh set face forgery score threshold is smaller than the third set face forgery score threshold, the eighth set face forgery score threshold is smaller than the fourth set face forgery score threshold, and the ninth set face forgery score threshold is smaller than the sixth set face forgery score threshold;
s1445, if the human eye region artifact detection result normalization score of the iris artifact type (expressed as) Greater than or equal to a first set iris artifact fraction threshold (which may be expressed as +.>) Determining the object to be detected as iris pseudo-object;
s1446, if the human eye region artifact detection result normalization score of the iris artifact type (expressed as) Less than the first set iris artifact fraction threshold (representable as +.>) And greater than or equal to a second set iris artifact fraction threshold (which may be expressed as +.>) And the multi-modal iris artifact detection result normalizes a score (expressed as ) Less than a third set iris artifact score threshold (which may be expressed as +.>) And greater than or equal to a fourth set iris artifact fraction threshold (which may be expressed as +.>) Determining the object to be detected as iris pseudo-object; the second set iris artifact score threshold is less than the first set iris artifact score threshold, the fourth set iris artifact score threshold is less than the third set iris artifact score threshold, and the third set iris artifact score threshold is less than the first set iris artifact score threshold.
In this embodiment, each detection result may be a classification score, and each detection result may be normalized to the same range, for example, between 0 and 1. The embodiment can combine various detection results to comprehensively obtain the iris living detection result, so that the detection result is less prone to omission and more accurate.
Further, in the steps S1441 to S1446, the third set face artifact count threshold (may be expressed as) May be the first set face forgery fraction threshold (expressed as +.>) A first set fractional multiple (may be denoted as x) of the fourth set face artifact fractionNumber threshold (representable as +. >) May be the second set face forgery fraction threshold (expressed as +.>) A first set fractional multiple of (may be denoted as x), the sixth set face artifact fraction threshold (denoted as +.>) May be the fifth set face forgery fraction threshold (expressed as +.>) A first set fractional multiple of (may be denoted as x); the seventh set face forgery fraction threshold (which may be expressed as +.>) May be the first set face forgery fraction threshold (expressed as +.>) The eighth set face forgery fraction threshold (which may be expressed as +.>) May be the second set face artifact count threshold (expressed as) A second set fractional multiple of (may be denoted as y), the ninth set face forgery fraction threshold (denoted as +.>) May be the fifth set face forgery fraction threshold (expressed as +.>) A second set fractional multiple of (may be denoted as y); the second set fractional multiple (may be denoted as y) is less than the first set fractional multiple (may be denoted as x); the second set iris artifact score threshold (which may be expressed as +.>) May be the first set iris artifact fraction threshold (may be expressed as +.>) A first set fractional multiple of (may be denoted as x) and a fourth set iris artifact fraction threshold (may be denoted as +. >) May be the third set iris artifact fraction threshold (may be expressed as +.>) A first set fractional multiple of (may be denoted as x). In a more specific embodiment, the first set fractional multiple and the second set fractional multiple may be set according to circumstances. In other embodiments, the multiple between different thresholds may be different. />
For better early warning, the iris biopsy method shown in fig. 1 may further include the steps of: s170, acquiring the body temperature of the object to be detected, and outputting body temperature alarm information or iris living body detection result that the object to be detected is a forgery under the condition that the body temperature of the object to be detected exceeds a set human body temperature threshold range.
The set human body temperature threshold range may be the range of the maximum human body temperature, such as 30-40 degrees. For example, the body temperature range used for quarantine detection during epidemic situation can be the range showing the disease-free state of human body, and is often smaller than the set threshold value range of human body temperature. The prosthesis judgment can be carried out by means of the temperature measuring module by outputting the body temperature alarm information. Outputting the body temperature alarm information can make personnel notice that suspected prosthesis possibly appears, and in order to eliminate the influence of the stability of the temperature measuring module, the final prosthesis detection result can be obtained by combining the follow-up steps. Or if the temperature measuring module has better stability, the detection result which is considered as the prosthesis can be directly output so as to improve the detection speed.
In the specific implementation, in the step S170, the step of obtaining the body temperature of the object to be detected may specifically include the steps of:
s171, carrying out face recognition on the object to be detected by utilizing the visible light image and/or the near infrared image to obtain a face key point positioning result;
s172, determining the face shielding type of the object to be detected according to the positioning result of the face key points; the face shielding type is that the face is not shielded, the mask is worn, the goggles are worn, and the mask and the goggles are worn;
s173, if the face shielding type of the object to be detected is that the face is not shielded, calculating the average temperature of the whole face area of the object to be detected according to the temperature distribution data of the object to be detected, and taking the average temperature as the body temperature of the object to be detected; wherein the temperature distribution data of the object to be detected corresponds to the thermal radiation image;
s174, if the face shielding type of the object to be detected is shielding of a wearing mask, calculating the average temperature of a face area above the mask worn by the object to be detected according to the temperature distribution data of the object to be detected, and taking the average temperature as the body temperature of the object to be detected;
S175, if the face shielding type of the object to be detected is shielding of wearing goggles, calculating the average temperature of a face area below the goggles worn by the object to be detected according to the temperature distribution data of the object to be detected, and taking the average temperature as the body temperature of the object to be detected;
and S176, if the face shielding type of the object to be detected is shielding comprising a wearing mask and goggles, obtaining the temperature of the exposed skin at the upper edge of the goggles worn by the object to be detected according to the temperature distribution data of the object to be detected, and taking the temperature as the body temperature of the object to be detected.
The steps S173 to S176 are modes for acquiring the body temperature of the subject to be detected for each of the face occlusion types in step S172. The shielding condition of the shielding object such as goggles and masks can be determined by performing the key point positioning and face detection in the steps S111 to S113. If no key point of the mouth and the nose is detected in the visible light image, the mask can be considered to be worn; for another example, in step S113, the temperature of the area corresponding to the mapped corner key point is low, and it is considered that goggles are worn. In addition, in the step S176, in the case of wearing the mask and goggles, other articles may be worn, and then a large area of face shielding may occur, and at this time, the temperature measurement can be performed by searching for the exposed skin of the face, for example, the temperature of the exposed skin can be obtained by searching for the area with the generally higher temperature on the face in the thermal radiation image.
In a further embodiment, the detection steps such as step S120, step S130, etc. may be performed only when the body temperature of the subject to be detected is normal.
For example, the step S120 may be a step of inputting a multi-modal combined face area image obtained from at least one of the visible light image and the near infrared image and the thermal radiation image into a multi-modal face living body detection model to obtain a multi-modal face artifact detection result, and the step may specifically include the steps of: inputting a multi-mode combined face area image obtained according to at least one of the visible light image and the near infrared image and the thermal radiation image into a multi-mode face living body detection model under the condition that the body temperature of the object to be detected does not exceed the set human body temperature threshold range, so as to obtain a multi-mode face forgery detection result;
for another example, in the step S130, the step of performing the living body detection on the human eye region image in the near infrared image by using the human eye region living body detection model to obtain a human eye region forgery detection result may specifically include the steps of: and under the condition that the body temperature of the object to be detected does not exceed the set human body temperature threshold range, performing living detection on a human eye region image in the near infrared image by using a human eye region living detection model to obtain a human eye region forgery detection result.
Based on the iris living body detection method shown in fig. 1, the embodiment of the invention also provides an iris recognition access control method. Fig. 3 is a flowchart of an iris recognition entrance guard control method according to an embodiment of the invention. Referring to fig. 3, the iris recognition access control method of the embodiments may include:
step S210: the iris living body detection result of the object to be detected is obtained by utilizing the iris living body detection method of any embodiment of the invention;
step S220: and controlling an entrance guard switch according to the iris living body detection result of the object to be detected.
In addition, based on the same inventive concept as the iris biopsy method shown in fig. 1, the embodiment of the invention further provides an iris biopsy device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor implements the steps of the iris biopsy method described in any one of the embodiments when executing the program.
Based on the iris living body detection device provided by the embodiment of the invention, the embodiment of the invention also provides an iris recognition access control system. Fig. 4 is a schematic structural diagram of an iris recognition gate inhibition control system according to an embodiment of the invention, referring to fig. 4, the iris recognition gate inhibition control system of the embodiments may include: the iris biopsy device according to any embodiment of the present invention comprises a visible light camera 310, a near infrared camera 320, a temperature measurement module 330.
Wherein the visible light camera 310 may be used to collect a visible light image of an object to be detected; the near-infrared camera 320 may be used to acquire near-infrared images of an object to be detected; the temperature measurement module 330 may be used to collect a thermal radiation image of the object to be detected; the iris living body detection device according to any embodiment of the present invention may be used for obtaining an iris living body detection result of an object to be detected according to a visible light image, a near infrared image, and a thermal radiation image of the object to be detected.
In addition, the embodiment of the invention also provides an iris recognition access control system, which comprises: the iris recognition access control system of any embodiment of the invention.
The embodiment of the invention also provides a computer readable storage medium, on which a computer program is stored, which when being executed by a processor, implements the steps of the iris living detection method according to any embodiment of the invention and the iris recognition access control method according to any embodiment of the invention.
It should be noted that the iris recognition gate inhibition control method, the iris living detection device, the iris recognition gate inhibition system and the computer readable storage medium are based on the same inventive concept as the iris living detection method shown in fig. 1 or implemented by the method, and the principles of solving the problems are similar, so that the implementation can be referred to each other, and the repetition is omitted.
In order that those skilled in the art will better understand the present invention, embodiments of the present invention will be described below with specific examples.
In a specific embodiment, an iris recognition entrance guard control method with temperature measurement and high reliability living body detection is provided, which not only has the functions of registering and recognizing common iris recognition equipment, but also can be added with a temperature measurement module to realize high-precision temperature measurement under the condition of complex face shielding, thereby realizing multi-mode living body detection based on a thermal radiation diagram, a visible light face image and a near infrared iris image and further realizing entrance guard control based on iris recognition.
Fig. 5 is a flowchart of an iris recognition entrance guard control method according to an embodiment of the invention. Referring to fig. 5, the iris recognition entrance guard control method may include the steps of:
(1) At the beginning, a thermal radiation image, a visible light face image and a near infrared iris image can be acquired, and the thermal radiation face image can be obtained through image alignment;
(2) Measuring the body temperature T of the object to be detected, judging whether the body temperature T is abnormal (the body temperature T is smaller than the lower threshold Th_min or the body temperature T is larger than the upper threshold Th_max), if so, considering the object to be detected as a prosthesis, or carrying out body temperature abnormality alarm;
(3) If the body temperature T is normal, the visible light image, the thermal radiation image and the near infrared image can be utilized for multi-mode living body detection;
(4) And (3) integrating the detection results of the step (3) to perform multi-mode living body detection result fusion, if the living body is judged not to be living body, performing prosthesis alarm, otherwise, performing iris registration or iris recognition, and performing access control.
In the step (1), the following steps: a) The face region can be detected in the visible light image of a large scene shot by the visible light face camera to obtain a visible light face image, and key points (such as eyes, nose, mouth corners, face circumferences and the like) of the face can be obtained through positioning; b) According to the face detection result of the visible light image, the near-infrared camera is controlled and regulated to acquire clear near-infrared images, the iris is detected on the near-infrared images to obtain near-infrared iris images, eye key points can be positioned and obtained to obtain near-infrared human eye images, a subsequent iris living detection flow based on the near-infrared images can be carried out according to the near-infrared iris images and the near-infrared human eye images, and an iris recognition flow can be carried out; c) The detected visible light image or the position of the face region in the near infrared image can be mapped onto the thermal radiation pattern acquired by the temperature measuring module according to the position relation between the near infrared camera and the temperature measuring module, so as to obtain a face thermal radiation image (or the face region called thermal radiation face image or thermal radiation image) for measuring the temperature of the face.
In the step (2), the face temperature can be calculated in a classified manner according to various shielding models, and temperature compensation is performed to realize stable face temperature measurement, so as to obtain the body temperature of the object to be measured.
In particular, to achieve high accuracy of facial thermometry, on the one hand, depends on the accuracy of the thermometry module and, on the other hand, on the thermometry position and distance. In the case of a face without occlusion, it is relatively easy to locate the facial key points. However, when the face shielding condition is complicated, for example, a mask, glasses, a hat, other shielding objects and the like are worn, the difficulty in positioning the temperature measuring position is significantly increased. Therefore, in this embodiment, a model or strategy for face detection and key point positioning for complex face conditions is constructed for multiple face shielding conditions, so as to accurately position key points such as eyes, nose, mouth corners, and face circumference on a face. The method for face recognition and key point positioning can be realized by adopting the existing method. In the detection process, the faces with various different shielding conditions can be divided into a plurality of categories.
For example, the mask may be classified into several categories, such as normal, mask worn, goggles worn, large area mask shielding, etc. In this way, in the temperature measurement process, different temperature measurement positions and strategies can be selected according to different shielding conditions. For example, the corresponding policies may include: under normal conditions (no face shielding), the average temperature of the whole face can be calculated as the body temperature of the human body; under the condition of wearing the mask, the average temperature of the upper half part of the face can be calculated and used as the body temperature of the human body; under the condition of wearing goggles, if a mask is provided, the average temperature of the parts above the goggles can be calculated and used as the body temperature of a human body; under the condition of wearing goggles, if a mask is not worn, the average temperature of other facial areas except for the goggles can be calculated and used as the body temperature of a human body; in addition, under the condition of shielding the large-area mask, the upper edge of the mask (such as goggles) can be used for searching the temperature measurement of the exposed skin to obtain the body temperature.
In order to match with an accurate temperature measurement strategy, accurate positioning of key points of glasses and goggles (such as an edge frame) can be performed, so that influence of the glasses and goggles on temperature measurement is avoided. The positioning method can be performed by adopting the same algorithm as the positioning of the key points of the human face.
In the step (3), a multi-modal living body detection is performed. Since the iris recognition device can be generally provided with a near infrared camera and a visible light camera, a temperature measuring module can be added in the iris recognition device. The multi-mode living body detection based on the multi-source image is realized by using three types of image sensors. The living body detection based on the characteristic fusion of the thermal radiation diagram, the visible light face image and the near infrared image has higher precision than the living body detection by utilizing the characteristics of a single mode. Face counterfeits, e.g. counterfeits of printing paper, masks, models, and artificial heat sourcesThe detection score of (2) may be recorded as S peri The detection score of iris artifacts such as pupil, eye prosthesis, vitreous eyeball, etc. can be recorded as S ocular Can be used for detecting the score S peri Sum score S ocular Normalization is performed, for example, to a score between 0 and 1, wherein a higher score may indicate a higher probability that the detection object is a forgery (or a forgery target).
The multi-modality living body detection method may include three parts: firstly, detecting forgery based on a thermal radiation pattern; secondly, fusing living body detection based on a characteristic layer of a thermal radiation face region and a visible light face; thirdly, living body detection based on the near infrared image iris region.
Specifically, in the process of performing forgery detection based on the heat radiation pattern: because the prosthesis generally has difficulty in achieving a realistic human body temperature effect, the heat radiation patterns of the face and the area near the face can be used for judging living bodies, and whether the prosthesis is a human face artifact or not can be detected and distinguished, such as whether the prosthesis is an unmanned body temperature prosthesis, an artificial heat source, a wearing mask or the like or not. The forgery detection score based on the thermal radiation image may be recorded asBecause the complexity of the heat radiation diagram is relatively low, the false heat source generally shows a more obvious difference from a real person, a lightweight depth network model can be used for classifying whether the false heat source is a false object or not, and the false heat source can be realized by using a depth convolution network.
In the living body detection process based on the fusion of the feature layers of the thermal radiation face region and the visible light face: the positions of the face camera (visible light camera) and the temperature measuring module are relatively easy to fix, and the face region in the visible light image and the face region in the thermal radiation diagram can be easily and accurately aligned, so that the two modes can be selected to be extracted for feature and analysis, and feature layer fusion learning and classification can be performed in a twin network mode (shown in fig. 6) or a multichannel input mode (pseudo twin network mode) (shown in fig. 7). Face forgery detection based on heat radiation face region and visible light face forgery detection, such as unmanned body temperature Prosthesis, artificial heat source, prosthesis attack wearing mask, etc., score may be recorded asDepending on the range and resolution of the images used for classification, if the iris areas in the visible light image and the thermal radiation image are large enough, both can be used for detection of iris artifacts such as wear pupil, artificial eye, glass eyeball, etc., the detection score can be recorded as +.>
In the living body detection process based on the near infrared image iris region: the detection of iris artifacts in near infrared images is easier than the detection of iris artifacts in visible images, and the detection of iris artifacts such as worn mydriasis, artificial eyes, glass eyeballs, and the like is also more accurate. Therefore, the present embodiment chooses to perform independent iris artifact detection by the iris region in the near infrared image, and finally can perform fractional fusion with the other two modalities. As shown in fig. 8, using the human eye region living body detection model, face artifacts (such as print paper, model, etc.) can be classified and detected based on the convolution network of the near infrared whole-eye image, and the classification score can be recorded asIn addition, the detection of iris artifacts (such as artifacts like wearing mydriasis, artificial eyes, glass eyeballs and the like) and the feature extraction of the whole periocular image can share the first three convolution network layers, the positioned iris region is intercepted on the convolution feature layers, and the convolution layers and the full convolution layers are connected to carry out iris artifact classification detection, and the classification score can be recorded as- >
In the step (4), the detection results of the three parts in the step (3) can be fused and calculated, and a simple fractional layer calculation fusion method can be adopted to ensureThe final classification score is determined. Fractional layer fusion is respectively aimed at S peri And S is ocular The calculation is performed as follows:
wherein alpha, beta, lambda, gamma, chi are fusion parameters, regulate the effect of different classifications on the final classification,normalized, <' > is performed>Normalization processing was performed. The classification parameters can be set manually according to classification results, and can also be determined in a learning mode.
Or whether the counterfeit attack behavior exists can be judged by adopting a strategy, and according to the actual condition of the system and the result of the living body detection of each part, a reasonably designed fusion strategy is used for the living body detection. The strategy design scheme is exemplified as follows:
(A) Judgment strategy for forgery of printing paper, model and the like
If it isOr->Then it is considered as a prosthesis (forgery of printing paper, model, etc.), wherein +.>Andsetting for larger thresholdSetting;
on the contrary, ifAnd->Or (I)>And->Then it is considered as a prosthesis (forgery of printing paper, model, etc.);
in the opposite direction, ifAnd->And->Then it is considered a prosthesis (forgery of printing paper, model, etc.).
(B) Judgment strategy for wearing artifacts such as pupil beauty, eye prosthesis and glass eyeball
If it isThen it is considered as a prosthesis (pseudo-object such as wearing a pupil, an artificial eye, a vitreous eyeball, etc.), wherein ∈ ->Setting for a larger threshold;
on the contrary, ifAnd->Then it is considered as a prosthesis (worn)Pseudomorphic objects such as pupil, eye prosthesis, vitreous eyeball, etc.).
(C) In the above processes (a) and (B), if one of the above processes is judged to be a prosthesis, the object to be detected is finally determined to be a prosthesis.
The embodiment provides a temperature measurement and living body detection method based on high reliability, which can provide an iris recognition access control system with temperature measurement and high reliability living body detection. The high-reliability living body detection of multi-mode fusion by adopting the thermal radiation image is realized, and the problems of body temperature measurement and living body detection under the condition of complex face shielding can be solved.
In summary, the iris living detection method, the iris recognition gate inhibition control method, the iris living detection device, the iris recognition gate inhibition control system, the iris recognition gate inhibition system and the computer readable storage medium according to the embodiments of the present invention can realize gate inhibition control management of high-precision iris recognition under complex face shielding by realizing highly reliable living detection using thermal radiation images for multi-modal fusion.
In the description of the present specification, reference to the terms "one embodiment," "one particular embodiment," "some embodiments," "for example," "an example," "a particular example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. The order of steps involved in the embodiments is illustrative of the practice of the invention, and is not limited and may be suitably modified as desired.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The foregoing description of the embodiments has been provided for the purpose of illustrating the general principles of the invention, and is not meant to limit the scope of the invention, but to limit the invention to the particular embodiments, and any modifications, equivalents, improvements, etc. that fall within the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (15)
1. An iris in-vivo detection method, comprising:
acquiring a thermal radiation image, a visible light image and a near infrared image of an object to be detected; the thermal radiation image, the visible light image and the face region in the near infrared image can be aligned;
inputting a multi-mode combined face area image obtained according to at least one of the visible light image and the near infrared image and the thermal radiation image into a multi-mode face living body detection model to obtain a multi-mode face forgery detection result for judging whether the object to be detected is a face forgery; the multi-mode human face living body detection model is obtained by training a first neural network with corresponding multiple inputs through a first training sample conforming to the multi-mode combination, and the set of the first training samples comprises training samples of a true human face type and a human face pseudo-object type;
Performing living body detection on a human eye region image in the near infrared image by using a human eye region living body detection model to obtain a human eye region forgery detection result; the human eye region living body detection model is obtained by training a second neural network by using a second training sample conforming to a human eye near infrared image mode, and the set of the second training samples comprises training samples of a true human eye type and a human eye region forgery type; the human eye region forgery type includes an iris forgery type, and the human eye region forgery detection result includes a human eye region forgery detection result for judging whether the object to be detected is an iris forgery type of an iris forgery;
obtaining an iris living body detection result of the object to be detected according to the multi-mode face forgery detection result and the human eye area forgery detection result, and if the object to be detected is judged to be a face forgery or an iris forgery according to the multi-mode face forgery detection result and the human eye area forgery detection result, confirming that the object to be detected does not pass through living body detection;
the first neural network is a multichannel input network according to the face area image of the multi-mode combination obtained by at least one of the visible light image and the near infrared image and the thermal radiation image;
Or,
according to the condition of the face area image of the multi-mode combination obtained by one of the visible light image and the near infrared image and the thermal radiation image, the first neural network is a twin network;
the second neural network comprises a first convolution network layer, a second convolution network layer and a second output layer;
the first convolution network layer is used for extracting characteristics of a human eye area image in the received near infrared image; the input end and the output end of the second convolution network layer are respectively connected with the output end and the second output layer of the first convolution network layer, and are used for carrying out feature extraction on iris region features obtained from feature extraction results of human eye region images in the near infrared images according to iris region positioning results in the human eye region images in the near infrared images, and outputting human eye region forgery detection results of iris forgery types according to feature extraction results of the iris region features.
2. The iris in-vivo detection method of claim 1, wherein acquiring a thermal radiation image, a visible light image, and a near infrared image of the object to be detected comprises:
Collecting a visible light image comprising the object to be detected, and detecting the position of a face area in the visible light image to obtain the face area image in the visible light image;
according to the position of a human face region in the visible light image, controlling and collecting a near infrared image comprising a clear iris of the object to be detected, obtaining a human face region image and a human eye region image in the near infrared image, and carrying out iris positioning on the human eye region image in the near infrared image to obtain an iris region image in the near infrared image;
and acquiring a thermal radiation image comprising the object to be detected, and mapping the face region image in the visible light image or the near infrared image to the thermal radiation image according to the position relation of the related camera to obtain the face region image in the thermal radiation image.
3. The iris in vivo detection method of claim 1, wherein,
the human eye region forgery type also comprises a human face forgery type, and the obtained human eye region forgery detection result also comprises a human eye region forgery detection result of the human face forgery type for judging whether the object to be detected is the human face forgery;
The second neural network further includes a first output layer;
the first output layer is connected with the output end of the first convolution network layer and is used for outputting human eye region forgery detection results of human face forgery types according to the feature extraction results of human eye region images in the near infrared images.
4. The iris in vivo detection method of claim 1, wherein,
the method further comprises the steps of:
performing living body detection on the thermal radiation image by using a human body temperature living body detection model to obtain a human face forgery detection result based on the human body temperature; the human body temperature living body detection model is obtained by training a third neural network by using a third training sample conforming to a thermal radiation image mode, and the set of the third training sample comprises training samples of a real human face type and a human face pseudo-object type;
obtaining an iris living body detection result of the object to be detected according to the multi-mode face forgery detection result and the human eye area forgery detection result, wherein the iris living body detection result comprises the following steps:
and obtaining an iris living body detection result of the object to be detected according to the multi-mode human face false body detection result, the human eye region false body detection result and the human face false body detection result based on the human body temperature.
5. The iris in vivo detection method of claim 1, wherein,
the method further comprises the steps of:
inputting a multi-mode combined iris region image obtained according to at least one of the visible light image and the near infrared image and the thermal radiation image into a multi-mode iris living body detection model to obtain a multi-mode iris forgery detection result; the multi-modal iris living detection model is obtained by training a fourth neural network with corresponding multi-input by using a fourth training sample conforming to the multi-modal combination, wherein the set of the fourth training samples comprises training samples of a true human iris type and an iris pseudo-object type, and the fourth neural network is the same as the first neural network;
obtaining an iris living body detection result of the object to be detected according to the multi-mode face forgery detection result and the human eye area forgery detection result, wherein the iris living body detection result comprises the following steps:
and obtaining an iris living body detection result of the object to be detected according to the multi-mode face forgery detection result, the human eye area forgery detection result and the multi-mode iris forgery detection result.
6. The iris in vivo detection method of claim 3, wherein,
The method further comprises the steps of:
performing living body detection on the thermal radiation image by using a human body temperature living body detection model to obtain a human face forgery detection result based on the human body temperature; the human body temperature living body detection model is obtained by training a third neural network by using a third training sample conforming to a thermal radiation image mode, and the set of the third training sample comprises training samples of a real human face type and a human face pseudo-object type; the method comprises the steps of,
inputting a multi-mode combined iris region image obtained according to at least one of the visible light image and the near infrared image and the thermal radiation image into a multi-mode iris living body detection model to obtain a multi-mode iris forgery detection result; the multi-modal iris living detection model is obtained by training a fourth neural network with corresponding multi-input by using a fourth training sample conforming to the multi-modal combination, wherein the set of the fourth training samples comprises training samples of a true human iris type and an iris pseudo-object type, and the fourth neural network is the same as the first neural network;
obtaining an iris living body detection result of the object to be detected according to the multi-mode face forgery detection result and the human eye area forgery detection result, wherein the iris living body detection result comprises the following steps:
Normalizing and weighting and summing the human face false object detection result based on the human body temperature, the multi-mode human face false object detection result and the human eye area false object detection result of the human face false object type to obtain a detection result that the object to be detected is the human face false object; wherein the facial artifact comprises one or more of an unmanned body temperature prosthesis, an artificial heat source body, and a worn mask, the unmanned body temperature prosthesis comprising one or more of a printed paper and a mannequin;
normalizing and weighting summing the multi-mode iris artifact detection result and the human eye region artifact detection result of the iris artifact type to obtain a detection result that the object to be detected is an iris artifact; iris artifacts include one or more of a pupil, an artificial eye, and a vitreous eyeball to be worn;
and outputting a detection result that the object to be detected is a human face artifact and/or a detection result that the object to be detected is an iris artifact as an iris living detection result of the object to be detected.
7. The iris in vivo detection method of claim 3, wherein,
the method further comprises the steps of:
Performing living body detection on the thermal radiation image by using a human body temperature living body detection model to obtain a human face forgery detection result based on the human body temperature; the human body temperature living body detection model is obtained by training a third neural network by using a third training sample conforming to a thermal radiation image mode, and the set of the third training sample comprises training samples of a real human face type and a human face pseudo-object type; the method comprises the steps of,
inputting a multi-mode combined iris region image obtained according to at least one of the visible light image and the near infrared image and the thermal radiation image into a multi-mode iris living body detection model to obtain a multi-mode iris forgery detection result; the multi-modal iris living detection model is obtained by training a fourth neural network with corresponding multi-input by using a fourth training sample conforming to the multi-modal combination, wherein the set of the fourth training samples comprises training samples of a true human iris type and an iris pseudo-object type, and the fourth neural network is the same as the first neural network;
obtaining an iris living body detection result of the object to be detected according to the multi-mode face forgery detection result and the human eye area forgery detection result, wherein the iris living body detection result comprises the following steps:
If the normalized score of the human face false artifact detection result based on the human body temperature is greater than or equal to a first set human face false artifact score threshold value or the normalized score of the human eye area false artifact detection result of the human face false artifact type is greater than or equal to a second set human face false artifact score threshold value, determining that the object to be detected is a human face false artifact;
if the normalized score of the human body temperature-based human face forgery detection result is smaller than the first set human face forgery score threshold and larger than or equal to a third set human face forgery score threshold, and the normalized score of the human eye region forgery detection result of the human face forgery type is smaller than the second set human face forgery score threshold and larger than or equal to a fourth set human face forgery score threshold, determining that the object to be detected is a human face forgery; the third set face forgery score threshold is smaller than the first set face forgery score threshold, and the fourth set face forgery score threshold is smaller than the second set face forgery score threshold;
if the normalized score of the human body temperature-based face forgery detection result is smaller than the first set face forgery score threshold and larger than or equal to the third set face forgery score threshold, and the normalized score of the multi-mode face forgery detection result is smaller than the fifth set face forgery score threshold and larger than or equal to the sixth set face forgery score threshold, determining that the object to be detected is a face forgery; wherein the fifth set face forgery score threshold is less than the first set face forgery score threshold and the second set face forgery score threshold, and the sixth set face forgery score threshold is less than the fifth set face forgery score threshold;
If the normalized score of the human body temperature-based human face forgery detection result is smaller than the third set human face forgery score threshold and larger than or equal to a seventh set human face forgery score threshold, and the normalized score of the human eye region forgery detection result of the human face forgery type is smaller than the fourth set human face forgery score threshold and larger than or equal to an eighth set human face forgery score threshold, and the normalized score of the multi-mode human face forgery detection result is smaller than the sixth set human face forgery score threshold and larger than or equal to a ninth set human face forgery score threshold, determining that the object to be detected is human face forgery; the seventh set face forgery score threshold is smaller than the third set face forgery score threshold, the eighth set face forgery score threshold is smaller than the fourth set face forgery score threshold, and the ninth set face forgery score threshold is smaller than the sixth set face forgery score threshold;
if the normalized score of the human eye region forgery detection result of the iris forgery type is larger than or equal to a first set iris forgery score threshold value, determining that the object to be detected is iris forgery;
If the normalized score of the human eye region false object detection result of the iris false object type is smaller than the first set iris false object score threshold and larger than or equal to the second set iris false object score threshold, and the normalized score of the multi-mode iris false object detection result is smaller than the third set iris false object score threshold and larger than or equal to the fourth set iris false object score threshold, determining that the object to be detected is iris false object; the second set iris artifact score threshold is less than the first set iris artifact score threshold, the fourth set iris artifact score threshold is less than the third set iris artifact score threshold, and the third set iris artifact score threshold is less than the first set iris artifact score threshold.
8. The iris in-vivo detection method of claim 7 wherein,
the third set face forgery score threshold is a first set fractional multiple of the first set face forgery score threshold, the fourth set face forgery score threshold is a first set fractional multiple of the second set face forgery score threshold, and the sixth set face forgery score threshold is a first set fractional multiple of the fifth set face forgery score threshold;
The seventh set face forgery score threshold is a second set fractional multiple of the first set face forgery score threshold, the eighth set face forgery score threshold is a second set fractional multiple of the second set face forgery score threshold, and the ninth set face forgery score threshold is a second set fractional multiple of the fifth set face forgery score threshold;
the second set fractional multiple is smaller than the first set fractional multiple;
the second set iris artifact score threshold is a first set fractional multiple of the first set iris artifact score threshold and the fourth set iris artifact score threshold is a first set fractional multiple of the third set iris artifact score threshold.
9. The iris in vivo detection method of claim 1, further comprising:
acquiring the body temperature of the object to be detected, and outputting body temperature alarm information or iris living body detection result that the object to be detected is forgery under the condition that the body temperature of the object to be detected exceeds a set human body temperature threshold range.
10. The iris biopsy method of claim 9, wherein the acquiring the body temperature of the subject to be tested, comprises:
Performing face recognition on the object to be detected by utilizing the visible light image and/or the near infrared image to obtain a face key point positioning result;
determining the face shielding type of the object to be detected according to the positioning result of the face key points; the face shielding type is that the face is not shielded, the mask is worn, the goggles are worn, and the mask and the goggles are worn;
if the face shielding type of the object to be detected is that the face is not shielded, calculating the average temperature of the whole face area of the object to be detected according to the temperature distribution data of the object to be detected, and taking the average temperature as the body temperature of the object to be detected; wherein the temperature distribution data of the object to be detected corresponds to the thermal radiation image;
if the face shielding type of the object to be detected is shielding of a wearing mask, calculating the average temperature of a face area above the mask worn by the object to be detected according to the temperature distribution data of the object to be detected, and taking the average temperature as the body temperature of the object to be detected;
if the face shielding type of the object to be detected is shielding of wearing goggles, calculating the average temperature of a face area below the goggles worn by the object to be detected according to the temperature distribution data of the object to be detected, and taking the average temperature as the body temperature of the object to be detected;
And if the face shielding type of the object to be detected is shielding comprising a wearing mask and goggles, obtaining the temperature of the exposed skin at the upper edge of the goggles worn by the object to be detected according to the temperature distribution data of the object to be detected, and taking the temperature as the body temperature of the object to be detected.
11. The iris in vivo detection method of claim 9, wherein,
inputting a multi-modal combined face region image obtained according to at least one of the visible light image and the near infrared image and the thermal radiation image into a multi-modal face living body detection model to obtain a multi-modal face forgery detection result, comprising:
inputting a multi-mode combined face area image obtained according to at least one of the visible light image and the near infrared image and the thermal radiation image into a multi-mode face living body detection model under the condition that the body temperature of the object to be detected does not exceed the set human body temperature threshold range, so as to obtain a multi-mode face forgery detection result;
performing living body detection on a human eye region image in the near infrared image by using a human eye region living body detection model to obtain a human eye region forgery detection result, wherein the method comprises the following steps:
And under the condition that the body temperature of the object to be detected does not exceed the set human body temperature threshold range, performing living detection on a human eye region image in the near infrared image by using a human eye region living detection model to obtain a human eye region forgery detection result.
12. An iris recognition access control method is characterized by comprising the following steps:
obtaining an iris biopsy result of the object to be tested using the iris biopsy method according to any one of claims 1 to 11;
and controlling an entrance guard switch according to the iris living body detection result of the object to be detected.
13. An iris biopsy apparatus comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the method according to any one of claims 1 to 11 when the program is executed.
14. An iris recognition access control system, comprising:
the visible light camera is used for collecting visible light images of the object to be detected;
the near infrared camera is used for collecting near infrared images of the object to be detected;
the temperature measuring module is used for collecting a thermal radiation image of an object to be detected;
The iris biopsy apparatus of claim 13, wherein the iris biopsy result of the object to be detected is obtained based on a visible light image, a near infrared image, and a thermal radiation image of the object to be detected.
15. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the steps of the method according to any one of claims 1 to 12.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010599660.1A CN111967296B (en) | 2020-06-28 | 2020-06-28 | Iris living body detection method, access control method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010599660.1A CN111967296B (en) | 2020-06-28 | 2020-06-28 | Iris living body detection method, access control method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111967296A CN111967296A (en) | 2020-11-20 |
| CN111967296B true CN111967296B (en) | 2023-12-05 |
Family
ID=73360891
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010599660.1A Active CN111967296B (en) | 2020-06-28 | 2020-06-28 | Iris living body detection method, access control method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111967296B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112394973B (en) * | 2020-11-23 | 2024-03-12 | 山东理工大学 | Multi-language code plagiarism detection method based on pseudo-twin network |
| CN112466024B (en) * | 2020-11-24 | 2021-07-16 | 上海卓繁信息技术股份有限公司 | Intelligent epidemic prevention system |
| CN112396125B (en) * | 2020-12-01 | 2022-11-18 | 中国第一汽车股份有限公司 | Classification method, device, equipment and storage medium for positioning test scenes |
| CN112800847B (en) * | 2020-12-30 | 2023-03-24 | 广州广电卓识智能科技有限公司 | Face acquisition source detection method, device, equipment and medium |
| CN113506391A (en) * | 2021-06-08 | 2021-10-15 | 上海欣能信息科技发展有限公司 | Electric power operation access system based on 5G communication |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101833646A (en) * | 2009-03-11 | 2010-09-15 | 中国科学院自动化研究所 | Living iris detection method |
| WO2013131407A1 (en) * | 2012-03-08 | 2013-09-12 | 无锡中科奥森科技有限公司 | Double verification face anti-counterfeiting method and device |
| CN104834901A (en) * | 2015-04-17 | 2015-08-12 | 北京海鑫科金高科技股份有限公司 | Binocular stereo vision-based human face detection method, device and system |
| CN105138996A (en) * | 2015-09-01 | 2015-12-09 | 北京上古视觉科技有限公司 | Iris identification system with living body detecting function |
| CN106204815A (en) * | 2016-06-23 | 2016-12-07 | 江西洪都航空工业集团有限责任公司 | A kind of gate control system based on human face detection and recognition |
| CN106372615A (en) * | 2016-09-19 | 2017-02-01 | 厦门中控生物识别信息技术有限公司 | Face anti-counterfeiting identification method and apparatus |
| CN107273812A (en) * | 2017-05-22 | 2017-10-20 | 西安交通大学 | A kind of living body iris method for anti-counterfeit for authentication |
| CN109543640A (en) * | 2018-11-29 | 2019-03-29 | 中国科学院重庆绿色智能技术研究院 | A kind of biopsy method based on image conversion |
| CN110008813A (en) * | 2019-01-24 | 2019-07-12 | 阿里巴巴集团控股有限公司 | Face identification method and system based on In vivo detection technology |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10042993B2 (en) * | 2010-11-02 | 2018-08-07 | Homayoon Beigi | Access control through multifactor authentication with multimodal biometrics |
| US20160189149A1 (en) * | 2014-12-30 | 2016-06-30 | Ebay Inc. | Biometric systems and methods for authentication and determination of group characteristics |
| TWI666935B (en) * | 2017-07-12 | 2019-07-21 | 謝基生 | A mini thermography for enhance nir captures images |
| US11657133B2 (en) * | 2017-12-13 | 2023-05-23 | Princeton Identity, Inc. | Systems and methods of multi-modal biometric analysis |
-
2020
- 2020-06-28 CN CN202010599660.1A patent/CN111967296B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101833646A (en) * | 2009-03-11 | 2010-09-15 | 中国科学院自动化研究所 | Living iris detection method |
| WO2013131407A1 (en) * | 2012-03-08 | 2013-09-12 | 无锡中科奥森科技有限公司 | Double verification face anti-counterfeiting method and device |
| CN104834901A (en) * | 2015-04-17 | 2015-08-12 | 北京海鑫科金高科技股份有限公司 | Binocular stereo vision-based human face detection method, device and system |
| CN105138996A (en) * | 2015-09-01 | 2015-12-09 | 北京上古视觉科技有限公司 | Iris identification system with living body detecting function |
| CN106204815A (en) * | 2016-06-23 | 2016-12-07 | 江西洪都航空工业集团有限责任公司 | A kind of gate control system based on human face detection and recognition |
| CN106372615A (en) * | 2016-09-19 | 2017-02-01 | 厦门中控生物识别信息技术有限公司 | Face anti-counterfeiting identification method and apparatus |
| CN107273812A (en) * | 2017-05-22 | 2017-10-20 | 西安交通大学 | A kind of living body iris method for anti-counterfeit for authentication |
| CN109543640A (en) * | 2018-11-29 | 2019-03-29 | 中国科学院重庆绿色智能技术研究院 | A kind of biopsy method based on image conversion |
| CN110008813A (en) * | 2019-01-24 | 2019-07-12 | 阿里巴巴集团控股有限公司 | Face identification method and system based on In vivo detection technology |
Non-Patent Citations (6)
| Title |
|---|
| Face–Iris Multimodal Biometric Identification System;Basma Ammour等;Electronics;第9卷(第1期);1-8 * |
| Multimodal Biometric Based on Fusion of Ridge Features with Minutiae Features and Face Features;Garg H 等;International Journal of Information System Modeling and Design;第11卷(第1期);37-57 * |
| The rise and regulation of thermal facial recognition technology during the COVID-19 pandemic;Meredith Van Natta等;doi: 10.1093/jlb/lsaa038;第7卷(第1期);1-9 * |
| 基于人脸识别的移动终端考勤系统设计;杨舒林;中国优秀硕士学位论文全文数据库 (信息科技辑)(第（2019）02期);I138-895 * |
| 基于卷积神经网络的多模态生物识别方法研究;杨林;中国优秀硕士学位论文全文数据库 (信息科技辑)(第（2019）09期);I138-789 * |
| 指纹静脉多模识别光学系统的设计;叶文炜;何丽鹏;林峰;;激光与光电子学进展(第09期);370-376 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111967296A (en) | 2020-11-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111967296B (en) | Iris living body detection method, access control method and device | |
| Wang et al. | Human visual system-based fundus image quality assessment of portable fundus camera photographs | |
| EP3373798B1 (en) | Method and system for classifying optic nerve head | |
| Adal et al. | An automated system for the detection and classification of retinal changes due to red lesions in longitudinal fundus images | |
| KR101167535B1 (en) | Method of authenticating individual and apparatus of authenticating individual used for the same | |
| Jafari et al. | Automatic detection of melanoma using broad extraction of features from digital images | |
| Aquino | Establishing the macular grading grid by means of fovea centre detection using anatomical-based and visual-based features | |
| EP3680794A1 (en) | Device and method for user authentication on basis of iris recognition | |
| KR20160115501A (en) | Method and Apparatus for acquiring a biometric information | |
| CN106096608A (en) | Chest temperature anomaly district's localization method and device | |
| Hernandez et al. | Early detection of Alzheimer's using digital image processing through iridology, an alternative method | |
| US20220198831A1 (en) | System for determining one or more characteristics of a user based on an image of their eye using an ar/vr headset | |
| Laddi et al. | Non-invasive jaundice detection using machine vision | |
| Noah Akande et al. | Implementation of a framework for healthy and diabetic retinopathy retinal image recognition | |
| KR102259462B1 (en) | Apparatus and method for evaluating visual field defect using eye-tracking and computer readable storage medium | |
| Raza et al. | Hybrid classifier based drusen detection in colored fundus images | |
| Akram et al. | Microaneurysm detection for early diagnosis of diabetic retinopathy | |
| CN110399786A (en) | A kind of noninductive recognition methods and system | |
| US10956735B1 (en) | System and method for determining a refractive error from red reflex images of eyes | |
| Bhardwaj et al. | Two-tier grading system for npdr severities of diabetic retinopathy in retinal fundus images | |
| Patel et al. | Glaucoma detection using mask region-based convolutional neural networks | |
| CN111374632B (en) | Retinopathy detection method, device and computer readable storage medium | |
| Gandhimathi et al. | Detection of neovascularization in proliferative diabetic retinopathy fundus images. | |
| Charoenpong et al. | Pupil extraction system for Nystagmus diagnosis by using K-mean clustering and Mahalanobis distance technique | |
| Liang et al. | Melasma image segmentation using extreme learning machine |
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 |