CN114743253B - Living body detection method and system based on distance characteristics of key points of adjacent faces - Google Patents
Living body detection method and system based on distance characteristics of key points of adjacent faces Download PDFInfo
- Publication number
- CN114743253B CN114743253B CN202210659261.9A CN202210659261A CN114743253B CN 114743253 B CN114743253 B CN 114743253B CN 202210659261 A CN202210659261 A CN 202210659261A CN 114743253 B CN114743253 B CN 114743253B
- Authority
- CN
- China
- Prior art keywords
- key points
- group
- topological graph
- preset key
- distance
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
Abstract
The invention discloses a living body detection method and a living body detection system based on adjacent face key point distance characteristics, wherein the method comprises the following steps: acquiring a plurality of standard face images of the same face, detecting three-dimensional coordinates of preset key points, numbering the preset key points, and forming the preset key points into a standard topological graph structure; forming a standard topological graph distance characteristic according to the distance between two preset key points of adjacent numbers in the standard topological graph structure; acquiring a two-dimensional face image, detecting three-dimensional coordinates of preset key points, numbering the preset key points, and forming the preset key points into a current topological graph structure; grouping the distance characteristics of the current topological graph according to the distance between two preset key points which are numbered adjacently in the structure of the current topological graph; and matching the current topological graph distance characteristic with the standard topological graph distance characteristic, and if the matching is successful, the two-dimensional face image comes from the living face. The method of the invention overcomes the defect that the living body detection by adopting visible light is affected by uneven illumination.
Description
Technical Field
The invention belongs to the technical field of face recognition, and particularly relates to a living body detection method and system based on adjacent face key point distance characteristics.
Background
With the development of the face recognition technology, the face recognition technology is not only commercially used in the aspects of railways, aviation, banks, face-brushing payment and the like, but also has important application in the field of security protection. The face recognition technology brings convenience to people and brings potential safety hazards, although a traditional face recognition system can recognize different faces, the face recognition system is difficult to judge whether the faces are living bodies or photos, for example, a bank card bound on a mobile phone is attacked by non-living body data such as photos, videos and the like.
Therefore, the in vivo detection technique is increasingly emphasized. In order to prevent cheating of photos, videos and the like, the living body detection is generally nested in a module in face detection and face recognition or verification and is used for verifying whether the living body detection is the user, and the real person identity verification is completed for high-security scenes such as finance, entrance guard and the like, so that the face recognition system can operate safely and stably. However, the existing in vivo detection technology still has some defects: the in-vivo detection technology for deep learning needs a large number of data sets as a training basis, and the data sets are uniformly distributed under different illumination conditions, otherwise, the same in-vivo or non-in-vivo judgment can be carried out under different illumination conditions; the interactive living judgment is not good in user experience through means of voice, limb actions and the like, and is easy to be attacked by recorded videos.
Disclosure of Invention
The invention aims to overcome one or more defects in the prior art and provides a living body detection method and system based on the distance characteristics of key points of adjacent faces.
The purpose of the invention is realized by the following technical scheme:
according to a first aspect of the present invention, a living body detection method based on a distance feature of key points of adjacent faces is provided. The living body detection method includes:
acquiring a plurality of standard face images of the same face, and respectively detecting three-dimensional coordinates of all preset key points in each standard face image;
numbering all preset key points in each standard face image according to a preset rule;
respectively normalizing all preset key points in each standard personnel image into a first three-dimensional model to form a standard topological graph structure;
respectively calculating the distance between two preset key points of adjacent numbers in the standard topological graph structure, and respectively taking a multidimensional distance vector consisting of all distances calculated in each standard topological graph structure as a standard topological graph distance feature;
acquiring a two-dimensional face image, and detecting three-dimensional coordinates of all preset key points in the two-dimensional face image;
numbering all preset key points in the two-dimensional face image according to a preset rule;
normalizing all preset key points in the two-dimensional face image into a second three-dimensional model to form a current topological graph structure, wherein the second three-dimensional model has the same size as the first three-dimensional model;
calculating the distance between two preset key points of adjacent numbers in the current topological graph structure, and taking a multidimensional distance vector consisting of all calculated distances in the current topological graph structure as the distance characteristic of the current topological graph;
and respectively matching the current topological graph distance features with all standard topological graph distance features, and if the matching value of the current topological graph distance features and any one standard topological graph distance feature is smaller than a threshold value, and the difference between each component in the current topological graph distance features and the corresponding component in the standard topological graph distance features is smaller than the threshold value, determining that the two-dimensional face image is from a living face.
Preferably, the plurality of standard face images include a front face image, a face image deviated to the left by a first angle, a face image deviated to the left by a second angle, a face image deviated to the right by the first angle, a face image deviated to the right by the second angle, a face image deviated to the first angle upwards, a face image deviated to the second angle upwards, a face image deviated to the first angle downwards and a face image deviated to the second angle downwards.
Preferably, the first angle has a value range of [ 0, 15 ], and the second angle has a value range of [ 15, 30 ].
Preferably, the preset key points include a first group of key points for forming a face contour, a second group of key points for forming a left eyebrow, a third group of key points for forming a right eyebrow, a fourth group of key points for forming a nose bridge, a fifth group of key points for forming a nostril, a sixth group of key points for forming a left eye, a seventh group of key points for forming a right eye, an eighth group of key points for forming an outer lip, and a ninth group of key points for forming an inner lip, all the preset key points in the first group of key points form a V-shaped structure, all the preset key points in the second group of key points form an inverted V-shaped structure, all the preset key points in the third group of key points form an inverted V-shaped structure, all the preset key points in the fourth group of key points form a vertical line structure, and all the preset key points in the fifth group of key points form a V-shaped structure, all preset key points in the sixth group of key points form an elliptical structure, all preset key points in the seventh group of key points form an elliptical structure, all preset key points in the eighth group of key points form an elliptical structure, and all preset key points in the ninth group of key points form an elliptical structure.
Preferably, the preset rule includes:
numbering all the preset key points in the first group of key points in sequence according to a counterclockwise sequence by taking the leftmost preset key point in the first group of key points as a starting point;
numbering all the preset key points in the second group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the second group of key points as a starting point;
numbering all the preset key points in the third group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the third group of key points as a starting point;
numbering all the preset key points in the fourth group of key points in sequence from top to bottom by taking the leftmost preset key point in the fourth group of key points as a starting point;
numbering all the preset key points in the fifth group of key points in sequence according to a counterclockwise sequence by taking the leftmost preset key point in the fifth group of key points as a starting point;
numbering all the preset key points in the sixth group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the sixth group of key points as a starting point;
numbering all the preset key points in the seventh group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the seventh group of key points as a starting point;
taking the leftmost preset key point in the eighth group of key points as a starting point, and numbering all the preset key points in the eighth group of key points in sequence according to a clockwise sequence;
and numbering all the preset key points in the ninth group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the ninth group of key points as a starting point.
Preferably, the number of the preset key points in the first group of key points is 17, the number of the preset key points in the second group of key points is 5, the number of the preset key points in the third group of key points is 5, the number of the preset key points in the fourth group of key points is 4, the number of the preset key points in the fifth group of key points is 5, the number of the preset key points in the sixth group of key points is 6, the number of the preset key points in the seventh group of key points is 6, the number of the preset key points in the eighth group of key points is 12, and the number of the preset key points in the ninth group of key points is 8.
Preferably, the acquiring a two-dimensional face image and detecting three-dimensional coordinates of all preset key points in the two-dimensional face image includes:
acquiring a two-dimensional face image through a binocular camera;
detecting all preset key points in the two-dimensional face image;
and calculating the three-dimensional coordinates of all preset key points in the two-dimensional face image according to the internal parameters of the binocular camera.
Preferably, the calculation formula of the matching value of the current topological graph distance feature and the standard topological graph distance feature is as follows:
in the formula (I), the compound is shown in the specification,
representing the matching value of the distance feature of the front topological graph and the distance feature of the kth standard topological graph, f representing the distance feature of the current topological graph,
the kth standard topological graph distance feature is represented, and m represents the total number of standard topological graph distance features.
Preferably, the calculation formula of the difference between each component in the current topological graph distance feature and the corresponding component in the standard topological graph distance feature is as follows:
in the formula (I), the compound is shown in the specification,
representing the difference of the jth component of the current topology distance feature and the jth component of the kth standard topology distance feature,
the jth component representing the distance characteristic of the current topology,
and representing the jth component of the kth standard topological graph distance feature, n representing the number of preset key points in a standard face image, and m representing the total number of the standard topological graph distance features.
According to a second aspect of the present invention, a living body detection system based on the distance feature of the key points of the adjacent human faces is provided. The living body detecting system includes:
the face image acquisition module is used for acquiring a standard face image and a two-dimensional face image;
the key point detection module is used for detecting three-dimensional coordinates of all preset key points in the standard face image and the two-dimensional face image and numbering the preset key points;
the distance feature calculation module is used for normalizing all preset key points in the standard personnel image into the first three-dimensional model to form a standard topological graph structure, calculating the distance between two adjacent numbered preset key points in the standard topological graph structure, and taking a multi-dimensional distance vector formed by all distances calculated in the standard topological graph structure as a standard topological graph distance feature; the two-dimensional topological graph distance feature extraction module is used for normalizing all preset key points in the two-dimensional face image into a second three-dimensional model to form a current topological graph structure, calculating the distance between two adjacent numbered preset key points in the current topological graph structure, and taking a multi-dimensional distance vector formed by all calculated distances in the current topological graph structure as a current topological graph distance feature, wherein the second three-dimensional model and the first three-dimensional model are the same in size;
and the matching module is used for matching the current topological graph distance features with all the standard topological graph distance features respectively, and if the matching value of the current topological graph distance features and any one standard topological graph distance feature is smaller than a threshold value and the difference between each component in the current topological graph distance features and the corresponding component in the standard topological graph distance features is smaller than the threshold value, the two-dimensional face image is considered to be from a living body face.
The invention has the beneficial effects that:
(1) after the key points of the face image are detected, the method does not need deep learning training, but judges whether the face image is a living body or not according to the distance information between the key points, can more directly judge whether a target is non-living body data such as videos, electronic photos or printed photos and the like, and overcomes the defect that the living body detection is influenced by uneven illumination when visible light is adopted;
(2) the face image in the invention adopts a two-dimensional image, the precision of extracting the preset key points from the two-dimensional face image is higher than that of extracting the preset key points from the three-dimensional face image, and simultaneously, the speed of extracting the preset key points from the two-dimensional face image is also higher than that of extracting the preset key points from the three-dimensional face image.
Drawings
FIG. 1 is a flowchart of an embodiment of a living body detection method based on a distance feature of key points of adjacent faces according to the present invention;
FIG. 2 is a schematic diagram of a number of a preset key point in a standard face image or a two-dimensional face image;
FIG. 3 is a block diagram of an embodiment of a system for detecting a living body based on a distance feature of a key point of an adjacent face according to the present invention.
Detailed Description
The technical solutions of the present invention will be described clearly and completely with reference to the following embodiments, and it should be understood that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without inventive effort based on the embodiments of the present invention, are within the scope of the present invention.
Referring to fig. 1 to fig. 3, the present embodiment provides a living body detection method and system based on the distance feature of key points of adjacent faces:
one embodiment of the living body detection method based on the distance characteristics of the key points of the adjacent human faces, which is provided by the invention, comprises the following steps: as shown in fig. 1, a living body detection method based on the distance feature of key points of adjacent faces includes:
s100, multiple standard face images of the same face are obtained, and three-dimensional coordinates of all preset key points in each standard face image are detected respectively.
In yet another embodiment, the plurality of standard facial images include a front facial image, a facial image that is deviated to the left by a first angle, a facial image that is deviated to the left by a second angle, a facial image that is deviated to the right by a first angle, a facial image that is deviated to the right by a second angle, a facial image that is deviated to the first angle upwards, a facial image that is deviated to the second angle upwards, a facial image that is deviated to the first angle downwards, and a facial image that is deviated to the second angle downwards. That is, a frontal face image, a face image whose face deviates to the left by a first angle, a face image whose face deviates to the left by a second angle, a face image whose face deviates to the right by the first angle, a face image whose face deviates to the right by the second angle, a face image whose face deviates to the first angle upwards, a face image whose face deviates to the second angle upwards, a face image whose face deviates to the first angle downwards, and a face image whose face deviates to the second angle downwards are collected. Generally, the first angle has a value range of [ 0, 15 ], and the second angle has a value range of [ 15, 30 ], that is, the first angle is 0 to 15 degrees (excluding 15 degrees), and the second angle is 15 to 30 degrees (including 15 degrees).
In yet another embodiment, the preset key points include a first group of key points for constituting a face contour, a second group of key points for constituting a left eyebrow, a third group of key points for constituting a right eyebrow, a fourth group of key points for constituting a nose bridge, a fifth group of key points for constituting a nostril, a sixth group of key points for constituting a left eye, a seventh group of key points for constituting a right eye, an eighth group of key points for constituting an outer lip, and a ninth group of key points for constituting an inner lip, all the preset key points in the first group of key points constitute a V-shaped structure, all the preset key points in the second group of key points constitute an inverted V-shaped structure, all the preset key points in the third group of key points constitute an inverted V-shaped structure, all the preset key points in the fourth group of key points constitute a vertical line structure, and all the preset key points in the fifth group of key points constitute a V-shaped structure, all preset key points in the sixth group of key points form an elliptical structure, all preset key points in the seventh group of key points form an elliptical structure, all preset key points in the eighth group of key points form an elliptical structure, and all preset key points in the ninth group of key points form an elliptical structure. Generally, there are 68 preset keywords, where the number of the preset keywords in the first group of the keywords is 17, the number of the preset keywords in the second group of the keywords is 5, the number of the preset keywords in the third group of the keywords is 5, the number of the preset keywords in the fourth group of the keywords is 4, the number of the preset keywords in the fifth group of the keywords is 5, the number of the preset keywords in the sixth group of the keywords is 6, the number of the preset keywords in the seventh group of the keywords is 6, the number of the preset keywords in the eighth group of the keywords is 12, and the number of the preset keywords in the ninth group of the keywords is 8.
S200, numbering all preset key points in each standard face image according to preset rules.
In yet another embodiment, the preset rules include: numbering all the preset key points in the first group of key points in sequence according to a counterclockwise sequence by taking the leftmost preset key point in the first group of key points as a starting point; numbering all the preset key points in the second group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the second group of key points as a starting point; numbering all the preset key points in the third group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the third group of key points as a starting point; numbering all the preset key points in the fourth group of key points in sequence from top to bottom by taking the leftmost preset key point in the fourth group of key points as a starting point; numbering all the preset key points in the fifth group of key points in sequence according to a counterclockwise sequence by taking the leftmost preset key point in the fifth group of key points as a starting point; numbering all the preset key points in the sixth group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the sixth group of key points as a starting point; numbering all the preset key points in the seventh group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the seventh group of key points as a starting point; numbering all the preset key points in the eighth group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the eighth group of key points as a starting point; and numbering all the preset key points in the ninth group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the ninth group of key points as a starting point.
S300, normalizing all preset key points in each standard person image to a first three-dimensional model to form a standard topological graph structure.
S400, respectively calculating the distance between two preset key points of adjacent numbers in the standard topological graph structure, and respectively using a multi-dimensional distance vector formed by all distances calculated in each standard topological graph structure as a standard topological graph distance feature. For example, if the number of the preset key points in each standard face image is 68, a 67-dimensional distance vector composed of all distances calculated in one standard topological graph structure is used as the standard topological graph distance feature corresponding to the standard topological graph structure.
Fig. 2 is an example of preset key points, where 1 to 68 in fig. 2 represent 68 preset key points, where No. 0 to 16 are a first group of key points for constituting a face contour, No. 17 to 21 are a second group of key points for constituting a left eyebrow, No. 22 to 26 are a third group of key points for constituting a right eyebrow, No. 27 to 30 are a fourth group of key points for constituting a nose bridge, No. 31 to 35 are a fifth group of key points for constituting a nostril, No. 36 to 41 are a sixth group of key points for constituting a left eye, No. 42 to 47 are a seventh group of key points for constituting a right eye, No. 48 to 59 are an eighth group of key points for constituting an outer lip, and No. 60 to 67 are a ninth group of key points for constituting an inner lip.
The calculation formula of the distance between two adjacent numbered preset key points in fig. 2 is:
in the formula, w is the width of the first three-dimensional model after normalization in the x-axis direction, h is the height of the first three-dimensional model after normalization in the y-axis direction, and c is the dimension of the first three-dimensional model after normalization in the z-axis direction;
representing the three-dimensional coordinates of the preset keypoints numbered i before normalization,
representing the three-dimensional coordinates of the preset key points with the numbers i after normalization, i representing the numbers of the preset key points,
to represent
In which the value at which i is 0,
to represent
Where i is the value at 16,
to represent
Where i is the value at 8,
to represent
Where i is the value at 19,
to represent
In which the value at which i is 0,
to represent
In which the value at which i is 0,
is indicated by the reference number
And
is between two predetermined key pointsThe distance between the two parts is equal to the distance between the two parts,
denotes the normalized number of
The three-dimensional coordinates of the preset key points,
denotes the normalized number of
The three-dimensional coordinates of the preset key points.
S500, collecting a two-dimensional face image, and detecting three-dimensional coordinates of all preset key points in the two-dimensional face image.
In another embodiment, acquiring a two-dimensional face image, and detecting three-dimensional coordinates of all preset key points in the two-dimensional face image includes:
and S510, acquiring a two-dimensional face image through a binocular camera.
And S520, detecting all preset key points in the two-dimensional face image.
S530, calculating three-dimensional coordinates of all preset key points in the two-dimensional face image according to internal parameters (parallax, focal length and the like) of the binocular camera.
S600, numbering all preset key points in the two-dimensional face image according to a preset rule.
S700, normalizing all preset key points in the two-dimensional face image into a second three-dimensional model to form a current topological graph structure, wherein the size of the second three-dimensional model is the same as that of the first three-dimensional model.
S800, calculating the distance between two preset key points of adjacent numbers in the current topological graph structure, and taking a multidimensional distance vector formed by all distances calculated in the current topological graph structure as the distance feature of the current topological graph.
And S900, respectively matching the current topological graph distance features with all standard topological graph distance features, if the matching value of the current topological graph distance features and any one standard topological graph distance feature is smaller than a threshold value, and the difference between each component in the current topological graph distance features and the corresponding component in the standard topological graph distance features is smaller than the threshold value, determining that the two-dimensional face image is from a living body face, otherwise, determining that the two-dimensional face image is from a non-living body face.
In one embodiment, the matching value is a cosine similarity between the distance feature of the current topological graph and the distance feature of the standard topological graph, and the calculation formula of the matching value is as follows:
in the formula (I), the compound is shown in the specification,
representing the matching value of the distance feature of the front topological graph and the distance feature of the kth standard topological graph, f representing the distance feature of the current topological graph,
the kth standard topological graph distance feature is represented, and m represents the total number of standard topological graph distance features.
In one embodiment, the difference between each component in the current topological graph distance feature and the corresponding component in the standard topological graph distance feature is calculated by the formula:
in the formula (I), the compound is shown in the specification,
representing the difference of the jth component of the current topology distance feature and the jth component of the kth standard topology distance feature,
representing a current topologyThe jth component of the distance feature,
and representing the jth component of the kth standard topological graph distance feature, n representing the number of preset key points in a standard face image, and m representing the total number of the standard topological graph distance features.
One embodiment of the living body detection system based on the distance characteristics of the key points of the adjacent human faces provided by the invention comprises the following steps: as shown in fig. 3, a living body detection system based on the distance feature of key points of adjacent faces includes a face image acquisition module, a key point detection module, a distance feature calculation module, and a matching module.
The face image acquisition module is used for acquiring a standard face image and a two-dimensional face image. A binocular camera is adopted when the face image acquisition module acquires a two-dimensional face image; the face image acquisition module acquires images of a face at multiple angles when acquiring a standard face image, for example, a front face image, a face image of the face deviating to a first angle to the left, a face image of the face deviating to a second angle to the left, a face image of the face deviating to the first angle to the right, a face image of the face deviating to the second angle to the right, a face image of the face deviating to the first angle to the upper side, a face image of the face deviating to the second angle to the upper side, a face image of the face deviating to the first angle to the lower side, and a face image of the face deviating to the second angle to the lower side. Typically, the first angle is 0-15 degrees (excluding 15 degrees) and the second angle is 15-30 degrees (including 15 degrees).
The key point detection module is used for detecting three-dimensional coordinates of all preset key points in the standard face image and the two-dimensional face image and numbering the preset key points.
In yet another embodiment, the preset key points include a first group of key points for constituting a face contour, a second group of key points for constituting a left eyebrow, a third group of key points for constituting a right eyebrow, a fourth group of key points for constituting a nose bridge, a fifth group of key points for constituting a nostril, a sixth group of key points for constituting a left eye, a seventh group of key points for constituting a right eye, an eighth group of key points for constituting an outer lip, and a ninth group of key points for constituting an inner lip, all the preset key points in the first group of key points constitute a V-shaped structure, all the preset key points in the second group of key points constitute an inverted V-shaped structure, all the preset key points in the third group of key points constitute an inverted V-shaped structure, all the preset key points in the fourth group of key points constitute a vertical line structure, and all the preset key points in the fifth group of key points constitute a V-shaped structure, all preset key points in the sixth group of key points form an elliptical structure, all preset key points in the seventh group of key points form an elliptical structure, all preset key points in the eighth group of key points form an elliptical structure, and all preset key points in the ninth group of key points form an elliptical structure. Generally, the number of the preset keys is 68, where the number of the preset key points in the first group of key points is 17, the number of the preset key points in the second group of key points is 5, the number of the preset key points in the third group of key points is 5, the number of the preset key points in the fourth group of key points is 4, the number of the preset key points in the fifth group of key points is 5, the number of the preset key points in the sixth group of key points is 6, the number of the preset key points in the seventh group of key points is 6, the number of the preset key points in the eighth group of key points is 12, and the number of the preset key points in the ninth group of key points is 8.
In yet another embodiment, the preset rules include: numbering all the preset key points in the first group of key points in sequence according to a counterclockwise sequence by taking the leftmost preset key point in the first group of key points as a starting point; numbering all the preset key points in the second group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the second group of key points as a starting point; numbering all the preset key points in the third group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the third group of key points as a starting point; numbering all the preset key points in the fourth group of key points in sequence from top to bottom by taking the leftmost preset key point in the fourth group of key points as a starting point; numbering all the preset key points in the fifth group of key points in sequence according to a counterclockwise sequence by taking the leftmost preset key point in the fifth group of key points as a starting point; numbering all the preset key points in the sixth group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the sixth group of key points as a starting point; taking the leftmost preset key point in the seventh group of key points as a starting point, and numbering all the preset key points in the seventh group of key points in sequence according to a clockwise sequence; numbering all the preset key points in the eighth group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the eighth group of key points as a starting point; and numbering all the preset key points in the ninth group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the ninth group of key points as a starting point.
The distance feature calculation module is used for normalizing all preset key points in the standard personnel image into the first three-dimensional model to form a standard topological graph structure, calculating the distance between two adjacent numbered preset key points in the standard topological graph structure, and taking a multi-dimensional distance vector formed by all distances calculated in the standard topological graph structure as a standard topological graph distance feature; and the distance calculation module is used for normalizing all preset key points in the two-dimensional face image into a second three-dimensional model to form a current topological graph structure, calculating the distance between two adjacent numbered preset key points in the current topological graph structure, and taking a multi-dimensional distance vector formed by all distances calculated in the current topological graph structure as the distance feature of the current topological graph, wherein the size of the second three-dimensional model is the same as that of the first three-dimensional model.
The matching module is used for matching the current topological graph distance features with all standard topological graph distance features respectively, and if the matching value of the current topological graph distance features and any one standard topological graph distance feature is smaller than a threshold value, and the difference between each component in the current topological graph distance features and the corresponding component in the standard topological graph distance features is smaller than the threshold value, the two-dimensional face image is considered to be from a living body face.
The foregoing is illustrative of the preferred embodiments of the present invention, and it is to be understood that the invention is not limited to the precise form disclosed herein and is not to be construed as limited to the exclusion of other embodiments, and that various other combinations, modifications, and environments may be used and modifications may be made within the scope of the concepts described herein, either by the above teachings or the skill or knowledge of the relevant art. And that modifications and variations may be effected by those skilled in the art without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (10)
1. A living body detection method based on the distance characteristics of key points of adjacent faces is characterized by comprising the following steps:
acquiring a plurality of standard face images of the same face, and respectively detecting three-dimensional coordinates of all preset key points in each standard face image;
numbering all preset key points in each standard face image according to a preset rule;
respectively normalizing all preset key points in each standard personnel image into a first three-dimensional model to form a standard topological graph structure;
respectively calculating the distance between two preset key points of adjacent numbers in the standard topological graph structure, and respectively taking a multidimensional distance vector consisting of all distances calculated in each standard topological graph structure as a standard topological graph distance feature;
acquiring a two-dimensional face image, and detecting three-dimensional coordinates of all preset key points in the two-dimensional face image;
numbering all preset key points in the two-dimensional face image according to a preset rule;
normalizing all preset key points in the two-dimensional face image into a second three-dimensional model to form a current topological graph structure, wherein the second three-dimensional model has the same size as the first three-dimensional model;
calculating the distance between two preset key points of adjacent numbers in the current topological graph structure, and taking a multidimensional distance vector consisting of all calculated distances in the current topological graph structure as the distance characteristic of the current topological graph;
and respectively matching the current topological graph distance features with all standard topological graph distance features, and if the matching value of the current topological graph distance features and any one standard topological graph distance feature is smaller than a threshold value, and the difference between each component in the current topological graph distance features and the corresponding component in the standard topological graph distance features is smaller than the threshold value, determining that the two-dimensional face image is from a living face.
2. The in-vivo detection method based on the distance characteristics of the key points of the adjacent faces, according to claim 1, wherein the plurality of standard face images comprise face images, face images deviated to the left by a first angle, face images deviated to the left by a second angle, face images deviated to the right by the first angle, face images deviated to the right by the second angle, face images deviated to the first angle upwards, face images deviated to the second angle upwards, face images deviated to the first angle downwards and face images deviated to the second angle downwards.
3. The in-vivo detection method based on the distance characteristics of the key points of the adjacent faces as claimed in claim 2, wherein the value range of the first angle is [ 0, 15 ], and the value range of the second angle is [ 15, 30 ].
4. The in-vivo detection method based on the distance characteristics of the adjacent face key points according to claim 1, wherein the preset key points comprise a first group of key points for forming a face contour, a second group of key points for forming a left eyebrow, a third group of key points for forming a right eyebrow, a fourth group of key points for forming a nose bridge, a fifth group of key points for forming a nostril, a sixth group of key points for forming a left eye, a seventh group of key points for forming a right eye, an eighth group of key points for forming an outer lip, and a ninth group of key points for forming an inner lip, all the preset key points in the first group of key points form a V-shaped structure, all the preset key points in the second group of key points form an inverted V-shaped structure, all the preset key points in the third group of key points form an inverted V-shaped structure, and all the preset key points in the fourth group of key points form a vertical line structure, all preset key points in the fifth group of key points form a V-shaped structure, all preset key points in the sixth group of key points form an oval structure, all preset key points in the seventh group of key points form an oval structure, all preset key points in the eighth group of key points form an oval structure, and all preset key points in the ninth group of key points form an oval structure.
5. The in-vivo detection method based on the distance features of the key points of the adjacent faces as claimed in claim 4, wherein the preset rule comprises:
numbering all the preset key points in the first group of key points in sequence according to a counterclockwise sequence by taking the leftmost preset key point in the first group of key points as a starting point;
numbering all the preset key points in the second group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the second group of key points as a starting point;
numbering all the preset key points in the third group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the third group of key points as a starting point;
numbering all the preset key points in the fourth group of key points in sequence from top to bottom by taking the leftmost preset key point in the fourth group of key points as a starting point;
numbering all the preset key points in the fifth group of key points in sequence according to a counterclockwise sequence by taking the leftmost preset key point in the fifth group of key points as a starting point;
numbering all the preset key points in the sixth group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the sixth group of key points as a starting point;
numbering all the preset key points in the seventh group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the seventh group of key points as a starting point;
numbering all the preset key points in the eighth group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the eighth group of key points as a starting point;
and numbering all the preset key points in the ninth group of key points in sequence according to a clockwise sequence by taking the leftmost preset key point in the ninth group of key points as a starting point.
6. The in-vivo detection method based on the distance characteristics of the adjacent face key points as claimed in claim 4, wherein the number of preset key points in the first group of key points is 17, the number of preset key points in the second group of key points is 5, the number of preset key points in the third group of key points is 5, the number of preset key points in the fourth group of key points is 4, the number of preset key points in the fifth group of key points is 5, the number of preset key points in the sixth group of key points is 6, the number of preset key points in the seventh group of key points is 6, the number of preset key points in the eighth group of key points is 12, and the number of preset key points in the ninth group of key points is 8.
7. The in-vivo detection method based on the distance feature of the adjacent face key points as claimed in claim 1, wherein the step of collecting a two-dimensional face image and detecting the three-dimensional coordinates of all preset key points in the two-dimensional face image comprises:
acquiring a two-dimensional face image through a binocular camera;
detecting all preset key points in the two-dimensional face image;
and calculating the three-dimensional coordinates of all preset key points in the two-dimensional face image according to the internal parameters of the binocular camera.
8. The in-vivo detection method based on the distance features of the key points of the adjacent faces as claimed in claim 1, wherein the calculation formula of the matching value of the distance features of the current topological graph and the distance features of the standard topological graph is as follows:
in the formula (I), the compound is shown in the specification,
representing the matching value of the distance feature of the front topological graph and the distance feature of the kth standard topological graph, f representing the distance feature of the current topological graph,
the kth standard topological graph distance feature is represented, and m represents the total number of standard topological graph distance features.
9. The in-vivo detection method based on the distance feature of the key points of the adjacent human faces as claimed in claim 1, wherein the calculation formula of the difference between each component in the distance feature of the current topological graph and the corresponding component in the distance feature of the standard topological graph is as follows:
in the formula (I), the compound is shown in the specification,
representing the difference of the jth component of the current topology distance feature and the jth component of the kth standard topology distance feature,
the jth component representing the distance characteristic of the current topology,
and representing the jth component of the kth standard topological graph distance feature, n representing the number of preset key points in a standard face image, and m representing the total number of the standard topological graph distance features.
10. A living body detection system based on the distance characteristics of key points of adjacent faces is characterized by comprising the following steps:
the face image acquisition module is used for acquiring a standard face image and a two-dimensional face image;
the key point detection module is used for detecting three-dimensional coordinates of all preset key points in the standard face image and the two-dimensional face image and numbering the preset key points;
the distance feature calculation module is used for normalizing all preset key points in the standard personnel image into the first three-dimensional model to form a standard topological graph structure, calculating the distance between two adjacent numbered preset key points in the standard topological graph structure, and taking a multi-dimensional distance vector formed by all distances calculated in the standard topological graph structure as a standard topological graph distance feature; the two-dimensional topological graph distance feature extraction module is used for normalizing all preset key points in the two-dimensional face image into a second three-dimensional model to form a current topological graph structure, calculating the distance between two adjacent numbered preset key points in the current topological graph structure, and taking a multi-dimensional distance vector formed by all calculated distances in the current topological graph structure as a current topological graph distance feature, wherein the second three-dimensional model and the first three-dimensional model are the same in size;
and the matching module is used for matching the current topological graph distance features with all the standard topological graph distance features respectively, and if the matching value of the current topological graph distance features and any one standard topological graph distance feature is smaller than a threshold value and the difference between each component in the current topological graph distance features and the corresponding component in the standard topological graph distance features is smaller than the threshold value, the two-dimensional face image is considered to be from a living body face.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210659261.9A CN114743253B (en) | 2022-06-13 | 2022-06-13 | Living body detection method and system based on distance characteristics of key points of adjacent faces |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210659261.9A CN114743253B (en) | 2022-06-13 | 2022-06-13 | Living body detection method and system based on distance characteristics of key points of adjacent faces |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114743253A CN114743253A (en) | 2022-07-12 |
| CN114743253B true CN114743253B (en) | 2022-08-09 |
Family
ID=82286716
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210659261.9A Active CN114743253B (en) | 2022-06-13 | 2022-06-13 | Living body detection method and system based on distance characteristics of key points of adjacent faces |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114743253B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105205458A (en) * | 2015-09-16 | 2015-12-30 | 北京邮电大学 | Human face living detection method, device and system |
| CN105740779A (en) * | 2016-01-25 | 2016-07-06 | 北京天诚盛业科技有限公司 | Method and device for human face in-vivo detection |
| CN108549873A (en) * | 2018-04-19 | 2018-09-18 | 北京华捷艾米科技有限公司 | Three-dimensional face identification method and three-dimensional face recognition system |
| CN109117726A (en) * | 2018-07-10 | 2019-01-01 | 深圳超多维科技有限公司 | A kind of identification authentication method, device, system and storage medium |
| CN109271950A (en) * | 2018-09-28 | 2019-01-25 | 广州云从人工智能技术有限公司 | A kind of human face in-vivo detection method based on mobile phone forward sight camera |
| CN110059624A (en) * | 2019-04-18 | 2019-07-26 | 北京字节跳动网络技术有限公司 | Method and apparatus for detecting living body |
| CN110363067A (en) * | 2019-05-24 | 2019-10-22 | 深圳壹账通智能科技有限公司 | Auth method and device, electronic equipment and storage medium |
| CN110363111A (en) * | 2019-06-27 | 2019-10-22 | 平安科技（深圳）有限公司 | Human face in-vivo detection method, device and storage medium based on lens distortions principle |
| CN111652086A (en) * | 2020-05-15 | 2020-09-11 | 汉王科技股份有限公司 | Face living body detection method and device, electronic equipment and storage medium |
| CN112801038A (en) * | 2021-03-02 | 2021-05-14 | 重庆邮电大学 | Multi-view face living body detection method and system |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107220590B (en) * | 2017-04-24 | 2021-01-05 | 广东数相智能科技有限公司 | Anti-cheating network investigation method, device and system based on in-vivo detection |
| WO2018232717A1 (en) * | 2017-06-23 | 2018-12-27 | 中国科学院自动化研究所 | Method, storage and processing device for identifying authenticity of human face image based on perspective distortion characteristics |
-
2022
- 2022-06-13 CN CN202210659261.9A patent/CN114743253B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105205458A (en) * | 2015-09-16 | 2015-12-30 | 北京邮电大学 | Human face living detection method, device and system |
| CN105740779A (en) * | 2016-01-25 | 2016-07-06 | 北京天诚盛业科技有限公司 | Method and device for human face in-vivo detection |
| CN108549873A (en) * | 2018-04-19 | 2018-09-18 | 北京华捷艾米科技有限公司 | Three-dimensional face identification method and three-dimensional face recognition system |
| CN109117726A (en) * | 2018-07-10 | 2019-01-01 | 深圳超多维科技有限公司 | A kind of identification authentication method, device, system and storage medium |
| CN109271950A (en) * | 2018-09-28 | 2019-01-25 | 广州云从人工智能技术有限公司 | A kind of human face in-vivo detection method based on mobile phone forward sight camera |
| CN110059624A (en) * | 2019-04-18 | 2019-07-26 | 北京字节跳动网络技术有限公司 | Method and apparatus for detecting living body |
| CN110363067A (en) * | 2019-05-24 | 2019-10-22 | 深圳壹账通智能科技有限公司 | Auth method and device, electronic equipment and storage medium |
| CN110363111A (en) * | 2019-06-27 | 2019-10-22 | 平安科技（深圳）有限公司 | Human face in-vivo detection method, device and storage medium based on lens distortions principle |
| CN111652086A (en) * | 2020-05-15 | 2020-09-11 | 汉王科技股份有限公司 | Face living body detection method and device, electronic equipment and storage medium |
| CN112801038A (en) * | 2021-03-02 | 2021-05-14 | 重庆邮电大学 | Multi-view face living body detection method and system |
Non-Patent Citations (6)
| Title |
|---|
| Face Liveness Detection Benchmark based on Stereo Matching;Minghui Shi等;《ICDSC 2019: Proceedings of the 13th International Conference on Distributed Smart Cameras》;20190930;第1-6页 * |
| Robust face anti-spoofing with depth information;Yan Wang等;《J. Vis. Commun. Image R.》;20170919;第49卷;第332-337页 * |
| Three-dimensional and two-and-a-halfdimensional face recognition spoofing using three-dimensional printed models;Javier Galbally等;《IET Biometrics》;20160601;第83-91页 * |
| 基于人脸关键点的活体检测方法在门禁系统中的应用;乔冬;《中国优秀博硕士学位论文全文数据库(硕士) 工程科技Ⅱ辑》;20220315(第03期);第C038-2089页 * |
| 基于近红外与可见光双目视觉的活体人脸检测方法;邓茜文等;《计算机应用》;20200710;第40卷(第7期);第2096-2103页 * |
| 面向人脸识别的人脸活体检测方法研究;杨健伟;《中国优秀博硕士学位论文全文数据库(硕士) 信息科技辑》;20150415(第04期);第I138-1093页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114743253A (en) | 2022-07-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106203294B (en) | The testimony of a witness based on face character analysis unifies auth method | |
| Chen et al. | 3D free-form object recognition in range images using local surface patches | |
| CN103268497B (en) | A kind of human face posture detection method and the application in recognition of face | |
| CN105740780B (en) | Method and device for detecting living human face | |
| CN110070010A (en) | A kind of face character correlating method identified again based on pedestrian | |
| Vijayan et al. | Twins 3D face recognition challenge | |
| CN110008913A (en) | The pedestrian's recognition methods again merged based on Attitude estimation with viewpoint mechanism | |
| CN106407985B (en) | A kind of three-dimensional human head point cloud feature extracting method and its device | |
| CN105740781B (en) | Three-dimensional human face living body detection method and device | |
| CN111680588A (en) | Human face gate living body detection method based on visible light and infrared light | |
| CN110796101A (en) | Face recognition method and system of embedded platform | |
| CN106991390A (en) | A kind of hand-held testimony of a witness Compare System and method based on deep learning | |
| CN106570491A (en) | Robot intelligent interaction method and intelligent robot | |
| CN105740779A (en) | Method and device for human face in-vivo detection | |
| CN106355138A (en) | Face recognition method based on deep learning and key features extraction | |
| CN103049758A (en) | Method for realizing remote authentication by fusing gait flow images (GFI) and head and shoulder procrustes mean shapes (HS-PMS) | |
| CN102254165A (en) | Hand back vein identification method based on fusion of structural coding features and texture coding features | |
| CN105138967B (en) | Biopsy method and device based on human eye area active state | |
| Bhanu et al. | Human ear recognition by computer | |
| US20200065564A1 (en) | Method for determining pose and for identifying a three-dimensional view of a face | |
| CN106355139A (en) | Facial anti-fake method and device | |
| CN106203322A (en) | A kind of identity authorization system based on hand back vein and palmmprint fusion image and method | |
| CN107622246A (en) | Face identification method and Related product | |
| CN111862413A (en) | Method and system for realizing epidemic situation resistant non-contact multidimensional identity rapid identification | |
| CN106156739B (en) | A kind of certificate photo ear detection and extracting method based on face mask analysis |
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 |