CN110866442B - Real-time face recognition-based testimony-of-person integrated checking system and method - Google Patents
Real-time face recognition-based testimony-of-person integrated checking system and method Download PDFInfo
- Publication number
- CN110866442B CN110866442B CN201910944252.2A CN201910944252A CN110866442B CN 110866442 B CN110866442 B CN 110866442B CN 201910944252 A CN201910944252 A CN 201910944252A CN 110866442 B CN110866442 B CN 110866442B
- Authority
- CN
- China
- Prior art keywords
- dimensional
- face image
- dimensional face
- face
- image
- 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
- 238000000034 method Methods 0.000 title claims description 35
- 238000007689 inspection Methods 0.000 claims abstract description 19
- 238000013507 mapping Methods 0.000 claims description 48
- 238000005259 measurement Methods 0.000 claims description 47
- 238000004364 calculation method Methods 0.000 claims description 33
- 238000000605 extraction Methods 0.000 claims description 10
- 230000009467 reduction Effects 0.000 claims description 10
- 239000000284 extract Substances 0.000 claims description 3
- 238000005286 illumination Methods 0.000 abstract description 8
- 230000008859 change Effects 0.000 abstract description 5
- 230000014509 gene expression Effects 0.000 abstract description 5
- 230000008569 process Effects 0.000 description 14
- 230000036544 posture Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 230000009466 transformation Effects 0.000 description 3
- 235000014676 Phragmites communis Nutrition 0.000 description 2
- 238000011946 reduction process Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 210000003484 anatomy Anatomy 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/06—Topological mapping of higher dimensional structures onto lower dimensional surfaces
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
- G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/55—Depth or shape recovery from multiple images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/758—Involving statistics of pixels or of feature values, e.g. histogram matching
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
- G06T2207/30201—Face
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
Abstract
The application discloses a testimony of a witness inspection system based on real-time face identification, this system gathers in advance and types three-dimensional face image, and examine with three-dimensional face image as the foundation when testimony of a witness is examined, because three-dimensional face image can embody the face depth, can improve the adaptability of face image to illumination, gesture, expression, sheltering from, shooting angle change, make face forgery, video forgery become more difficult, consequently utilize three-dimensional face image can embody this characteristic of face depth, examine testimony of a witness with three-dimensional face image as the foundation, examined quality and rate of accuracy of examination system have been improved.
Description
Technical Field
The application relates to the technical field of identity verification, in particular to a system and a method for checking a testimony of a witness based on real-time face recognition.
Background
When transacting financial business, riding an airplane, passing customs and other affairs, it is common to check valid certificates such as identity cards, passports and the like and to check whether the licensee is the principal. The identity identification chips are arranged in the second-generation identity card and the new-version passport, the identity identification in the identity identification chip is read when the ID-card integrated examination is carried out, the registered face photo corresponding to the identity identification can be inquired in the identity database, and the comparison is carried out with the identity holder.
In the past, manual comparison is mainly adopted, and at present, real-time video pictures of a certificate holder are shot, and automatic comparison is carried out by using a face recognition technology, so that the speed is higher and the accuracy is higher.
At present, face photos registered in an identity database of a second-generation identity card and a new-version passport are two-dimensional face photos, and two-dimensional real-time video pictures of a licensee are shot for comparison and identification in the checking process.
Although the face recognition technology based on the two-dimensional image has achieved better recognition efficiency to a certain extent, the essence of the face recognition technology is a brief projection of a three-dimensional face model in a two-dimensional space, and the amount of information contained in the extracted model is greatly reduced, so that when the extracted face model is affected by factors such as expressions, postures and illumination, the recognition rate of a two-dimensional face recognition system may be greatly reduced, and the recognition accuracy and the recognition efficiency are affected
Disclosure of Invention
Object of the application
Based on this, in order to improve the identification accuracy and the identification efficiency during identity verification, eliminate the influence on the aspects of illumination, posture, expression, shielding, shooting angle and the like, which is caused by the identification accuracy of face identification, improve the difficulty of face counterfeiting and video counterfeiting, and finally improve the verification quality and accuracy, the application discloses the following technical scheme.
(II) technical scheme
In one aspect, the present application provides a testimony of a witness unification inspection system based on real-time face identification, includes:
the certificate reader is used for reading identity information contained in a certificate held by a certificate holder, identifying the identity information and extracting a pre-registered two-dimensional face image from an identity database according to an identification result;
the video information acquisition equipment is used for acquiring a three-dimensional face video picture of a licensee in real time, wherein the three-dimensional face video picture comprises a two-dimensional face video picture and corresponding face depth information;
the two-dimensional image comparison module is used for comparing the acquired two-dimensional face video picture with the pre-registered two-dimensional face image;
the three-dimensional face registration module is used for generating a three-dimensional face image based on the pre-registered two-dimensional face image and the corresponding face depth information and registering the three-dimensional face image and the pre-registered two-dimensional face image under the condition that the two-dimensional image comparison result shows that the two-dimensional image and the pre-registered two-dimensional face image meet the matching condition;
and the three-dimensional face checking module is used for extracting a corresponding three-dimensional face image from the identity database according to a recognition result obtained by recognizing the read identity information by the certificate reader when checking the certificate holder, comparing the three-dimensional face image with a three-dimensional face video image of the certificate holder acquired by the video information acquisition equipment during checking, and judging whether the certificate holder is the certificate owner according to a three-dimensional image comparison result.
In a possible embodiment, the two-dimensional image comparison module includes:
and the target extraction unit is used for extracting a face picture with the largest face frame area from the acquired two-dimensional face video pictures as a compared target image before comparing the acquired two-dimensional face video pictures with the pre-registered two-dimensional face images.
In a possible implementation manner, the two-dimensional image comparison module extracts feature points with representative faces from the acquired two-dimensional face video picture and the pre-registered two-dimensional face image respectively, and performs registration between the two-dimensional face video picture and the two-dimensional face image through matching of the feature points; and,
and the three-dimensional face registration module corresponds the face depth information to the two-dimensional face image according to the feature point corresponding relation between the two-dimensional face video picture and the two-dimensional face image after registration, so that the two-dimensional face image registered in the identity database is recorded as the three-dimensional face image.
In one possible embodiment, the three-dimensional face checking module includes:
the dimensionality reduction mapping unit is used for conformally mapping the face model of the three-dimensional face image to a two-dimensional plane circular area to obtain a first face image based on a Rich curvature stream, and conformally mapping the face model of the three-dimensional face video picture to the two-dimensional plane circular area to obtain a second face image;
an energy value extraction unit, configured to calculate an energy value generated by each triangular mesh vertex in the planar model of the first face image and the second face image in conformal calculation, respectively, and further obtain an energy minimum value among a plurality of energy values of each triangular mesh vertex;
and the chi-square statistic unit is used for generating an energy feature histogram according to the energy minimum value of each vertex of the first face image and the second face image, and judging the similarity between the first face image and the second face image by calculating and comparing chi-square statistic of the energy feature histogram so as to obtain the comparison result.
In one possible embodiment, the dimension reduction mapping unit includes:
the radius calculation subunit is used for calculating an initial Riemann measurement radius according to the side length of an initial triangular mesh of the face curved surface and the vertex of each triangular mesh;
the reverse distance calculation subunit is used for calculating the reverse distance of each grid edge according to the initial Riemannian measurement radius and the edge length of the corresponding triangular grid;
the curvature calculation subunit is used for calculating an inner angle of the triangular grid according to the inversion distance and calculating the discrete Gaussian curvature of each vertex according to the inner angle;
the measurement length calculation subunit is used for adjusting the radius of the circle according to the difference between the initial Gaussian curvature and the target curvature, further calculating the side length measured by the current triangular grid according to the new radius of the circle, the radius of the circle of the adjacent point and the reversal distance of the side where the two circle centers are located, and calculating the angle of each vertex;
and the traversal subunit is used for controlling all the subunits to traverse all the vertexes on the curved surface of the human face to obtain a planar measurement which is conformally equivalent to the triangular mesh measurement in the three-dimensional human face model, and embedding the planar measurement into a two-dimensional planar circular area according to the triangular mesh corner relationship to obtain a human face image after conformal mapping.
In one possible embodiment, the energy value extracting unit encodes the calculated energy values of the vertices of the triangular mesh, assigns weights to the encoded energy values, calculates a plurality of initial vertex energy values of the vertices of the triangular mesh according to the weights and different encoding orders, and sets a minimum value of the plurality of initial vertex energy values as an energy minimum value of the vertices of the triangular mesh.
On the other hand, the application also provides a testimony of a witness unification checking method based on real-time face recognition, which comprises the following steps:
reading identity information contained in a certificate held by a certificate holder, identifying, and extracting a pre-registered two-dimensional face image from an identity database according to an identification result;
acquiring a three-dimensional face video picture of a licensee in real time, wherein the three-dimensional face video picture comprises a two-dimensional face video picture and corresponding face depth information;
comparing the collected two-dimensional face video picture with the pre-registered two-dimensional face image;
under the condition that the two-dimensional image comparison result shows that the two images accord with the matching condition, generating a three-dimensional face image based on the pre-registered two-dimensional face image and the corresponding face depth information and registering;
when the certificate holder is checked, extracting a corresponding three-dimensional face image from the identity database according to a recognition result obtained by recognizing the read identity information, comparing the three-dimensional face image with a three-dimensional face video image of the certificate holder collected during checking, and judging whether the certificate holder is a certificate owner according to a three-dimensional image comparison result.
In a possible implementation manner, before the acquired two-dimensional face video picture is compared with the pre-registered two-dimensional face image, a face picture with a largest face frame area is extracted from the acquired two-dimensional face video picture as a comparison target image.
In a possible embodiment, the comparing the acquired two-dimensional face video picture with the pre-registered two-dimensional face image includes: extracting characteristic points with face representativeness from the collected two-dimensional face video picture and the pre-registered two-dimensional face image respectively, and registering the two-dimensional face video picture and the two-dimensional face image through matching of the characteristic points;
and the generating and registering a three-dimensional face image based on the pre-registered two-dimensional face image and the corresponding face depth information comprises: and according to the feature point correspondence between the two-dimensional face video picture and the two-dimensional face image after registration, corresponding the face depth information to the two-dimensional face image, thereby registering the two-dimensional face image registered in an identity database as the three-dimensional face image.
In a possible embodiment, the comparing the three-dimensional face image with the three-dimensional face video picture of the witness collected during the inspection includes:
based on a Rich curvature stream, conformally mapping a face model of the three-dimensional face image to a two-dimensional plane circular area to obtain a first face image, and conformally mapping the face model of the three-dimensional face video picture to the two-dimensional plane circular area to obtain a second face image;
respectively calculating energy values of vertexes of each triangular mesh in the plane models of the first face image and the second face image in conformal calculation, and further obtaining an energy minimum value in a plurality of energy values of the vertexes of each triangular mesh;
generating an energy feature histogram according to the energy minimum value of each vertex of the first face image and the second face image, and judging the similarity of the first face image and the second face image by calculating and comparing chi-square statistics of the energy feature histogram, so as to obtain the comparison result.
In one possible implementation, the conformal mapping of the face model of the face image to the two-dimensional planar circular region based on the stream of reed curvatures comprises:
calculating an initial Riemann measurement radius according to the side length of an initial triangular mesh of the face curved surface and the vertex of each triangular mesh;
calculating the reversal distance of each grid edge according to the initial Riemannian measurement radius and the edge length of the corresponding triangular grid;
calculating an inner angle of the triangular mesh according to the inversion distance, and calculating a discrete Gaussian curvature of each vertex according to the inner angle;
adjusting the radius of the circle according to the difference between the initial Gaussian curvature and the target curvature, further calculating the side length measured by the current triangular mesh according to the new radius of the circle, the radius of the circle of the adjacent point and the reversal distance of the side where the two circle centers are located, and calculating the angle of each vertex;
and controlling all the subunits to traverse all vertexes on the face curved surface to obtain plane measurement which is conformally equivalent to triangular mesh measurement in the three-dimensional face model, and embedding the plane measurement into a two-dimensional plane circular area according to the corner relationship of the triangular mesh to obtain a face image subjected to conformal mapping.
In a possible embodiment, the obtaining an energy minimum value of the plurality of energy values of each vertex of the triangular mesh includes:
and coding the calculated energy values of the triangular mesh vertexes, respectively distributing weights, calculating a plurality of initial vertex energy values of the triangular mesh vertexes according to the weights and different coding sequences, and taking the minimum value of the plurality of initial vertex energy values as the energy minimum value of the triangular mesh vertexes.
(III) advantageous effects
The application discloses a system and a method for checking the testimony of a witness based on real-time face recognition, wherein a three-dimensional face image is collected in advance and input, and the testimony of a witness is checked by taking the three-dimensional face image as a basis.
Drawings
The embodiments described below with reference to the drawings are exemplary and intended to be used for explaining and illustrating the present application and should not be construed as limiting the scope of the present application.
Fig. 1 is a block diagram of an embodiment of a testimonial-testimonial integrated inspection system disclosed in the present application.
Fig. 2 is a flowchart illustrating an embodiment of a testimonial to testimonial unification checking method disclosed in the present application.
Detailed Description
In order to make the implementation objects, technical solutions and advantages of the present application clearer, the technical solutions in the embodiments of the present application will be described in more detail below with reference to the drawings in the embodiments of the present application.
An embodiment of the witness-in-one inspection system disclosed in the present application is described in detail below with reference to fig. 1. As shown in fig. 1, the inspection system disclosed in this embodiment mainly includes: the system comprises a certificate reader, video information acquisition equipment, a two-dimensional image extraction module, a two-dimensional image comparison module, a three-dimensional face registration module and a three-dimensional face inspection module.
The certificate reader is used for reading identity information contained in a certificate held by a certificate holder and identifying the identity information, and extracting a pre-registered two-dimensional face image from an identity database according to an identification result.
The certificate reader can read identity information stored in a non-contact IC card chip embedded in an identity card or read information which can be used for identity identification and is stored in a micro radio frequency identification non-contact chip embedded in an electronic passport. The identity card and the electronic passport can store identity information including name, gender, date of birth, identification card number and the like. When the personal identity information contained in the certificate held by the certificate holder is checked, the identity card of the certificate holder can be placed in the card reading area of the certificate reader, and the identity information contained in the identity card can be read.
After the identity information of the certificate holder is read, the certificate reader searches in a pre-established identity database storing the identity information of each person, information of the person corresponding to the identity card is extracted, and the extracted information comprises a pre-acquired two-dimensional face image, such as a certificate photo of an identity card owner.
The pre-registration or pre-collection may refer to recording and saving the two-dimensional face image in a corresponding identity database. The pre-acquired two-dimensional face image may be referred to hereinafter as a first face image.
The video information acquisition equipment is used for acquiring three-dimensional face video pictures of the licensee in real time, and the three-dimensional face video pictures comprise two-dimensional face video pictures and corresponding face depth information.
Because the system of this embodiment is the integrated inspection system of testimony of a witness, consequently when examining the certifier identity correct, still can gather the people's face information of certifier through video information acquisition equipment. Different from the existing checking process, the video information acquisition equipment of the system is a three-dimensional face video acquisition device which acquires three-dimensional face video pictures of a licensee. The two-dimensional face video picture included in the three-dimensional face video picture may be referred to as a second face image in the following text, and the two-dimensional face video picture mainly includes a face image, even a picture that only includes a face image after image processing.
Three-dimensional face data has several advantages over two-dimensional face data:
1. the three-dimensional data is not influenced by illumination change because the three-dimensional data does not contain brightness information, and the influence of the posture of the real three-dimensional face data is far smaller than that of a two-dimensional image because the real three-dimensional face data has geometric and depth information;
2. because the human face is a three-dimensional non-rigid object in nature, compared with a two-dimensional human face image, the three-dimensional human face geometric anatomical structure has certain stability and robustness, and individuals have certain independence;
3. the anti-counterfeiting performance is far more than two-dimensional information due to the data expression mode, so that the time and economic cost for counterfeiting is far higher than two-dimensional, and deceptive means such as photos, makeup, videos and the like can be effectively overcome.
Specifically, this collection equipment can adopt the structured light principle, and it includes surveillance camera head, infrared structure light emitter and infrared camera, and two-dimentional people face video picture is gathered to the surveillance camera head, and infrared structure light emitter projects a plurality of light spots onto the person's head of holding a certificate, and infrared camera reads the dot matrix pattern and catches infrared image, gathers corresponding degree of depth information. The acquisition device may also employ a three-dimensional scanner, for example a tof (timeofflight) camera, which continuously emits a surface light source to the head of the bearer, and calculates distance information from the recorded time at which the reflected light reaches the receiver, to obtain depth information.
The two-dimensional image comparison module is used for comparing a two-dimensional face video picture contained in a three-dimensional face video picture collected by the video information collection equipment with a pre-registered two-dimensional face image extracted by the certificate reader.
Compared with the existing checking process, the two-dimensional image comparison module in the embodiment compares the acquired image with the registered image to judge whether the photo of the certificate holder is consistent with the certificate (identity card), and the purpose of the two-dimensional image comparison module is not only to judge whether the certificate holder is the certificate owner, but also to use the judgment result as the basis for judging whether the certificate holder can register the three-dimensional face image.
The three-dimensional face registration module is used for generating a three-dimensional face image based on a pre-registered two-dimensional face image extracted by the certificate reader and face depth information acquired by the video information acquisition equipment under the condition that a two-dimensional image comparison result of the two-dimensional image comparison module shows that the two-dimensional face video picture and the two-dimensional face image accord with a matching condition (namely the two-dimensional face video picture and the two-dimensional face image are matched), registering the generated three-dimensional face image, and taking the three-dimensional face image as a template for checking the face of the person.
The two-dimensional image comparison module judges that the certificate holder is the certificate owner through the two-dimensional image, and shows that the certificate holder can register the three-dimensional face image so as to be used for three-dimensional face recognition during later inspection. Since the two-dimensional face image (first face image) is a face image registered in advance, the two-dimensional face image is used as a basis for generating a three-dimensional face image. And because the two-dimensional face video picture (the second face image) that video information acquisition equipment gathered at present can match with two-dimensional face image, therefore the corresponding face depth information that video information acquisition equipment gathered also can be fit for with two-dimensional face image. Therefore, the face depth information is superimposed on the two-dimensional face image (first face image) to form a three-dimensional face image or three-dimensional face feature data.
Through the comparison of the two-dimensional image comparison module, the formed three-dimensional face image is ensured to be a correct three-dimensional image of the certificate owner, and the three-dimensional image also comprises the depth information of the face. Under the condition that the true and correct identity of the licensee is ensured through the two-dimensional image comparison module in advance, the correct three-dimensional face image of the licensee is collected and registered through the three-dimensional face registration module and is recorded into the identity database, and the additional recording of the three-dimensional face is achieved.
The three-dimensional face checking module is used for extracting a corresponding three-dimensional face image from the identity database according to a recognition result obtained by recognizing the read identity information by the certificate reader when checking the certificate holder, comparing the three-dimensional face image with a three-dimensional face video image of the certificate holder acquired by the video information acquisition equipment during checking, and judging whether the certificate holder is the certificate owner according to a three-dimensional image comparison result.
After the three-dimensional face is additionally recorded, the personnel identity checking process is carried out again: identity information of a certificate holder is read and identified through a certificate reader, then a three-dimensional face image corresponding to the identity information contained in the certificate is extracted, a three-dimensional face video picture of the certificate holder is collected through video information collection equipment, and then the three-dimensional face video picture and the three-dimensional face video picture are compared. If the comparison result shows that the identity of the licensee is consistent with the identity of the licensee, otherwise, the identity of the licensee is inconsistent.
The inspection system disclosed by the embodiment collects and inputs the three-dimensional face image in advance, and inspects the three-dimensional face image as the basis during the testimony inspection, and the three-dimensional face image can embody the face depth, so that the adaptability of the face image to the changes of illumination, posture, expression, shielding and shooting angles can be improved, the face counterfeiting and the video counterfeiting become more difficult, the testimony inspection is performed by utilizing the characteristic that the three-dimensional face image can embody the face depth, and the testimony inspection is performed by taking the three-dimensional face image as the basis, so that the inspection quality and the accuracy of the inspection system are improved.
Since a plurality of faces may appear in the video image captured by the video information capturing device, in order to implement one-to-one comparison of faces, in one embodiment, the two-dimensional image comparison module includes a target extraction unit, configured to extract a face image with a largest face frame area from the captured two-dimensional face video image as a comparison target image before comparing the captured two-dimensional face video image with a pre-registered two-dimensional face image.
In one embodiment, the two-dimensional image comparison module extracts feature points with face representativeness from the acquired two-dimensional face video picture and the pre-registered two-dimensional face image respectively, and performs registration between the two-dimensional face video picture and the two-dimensional face image through matching of the feature points.
The image registration method can adopt a gray information based method, a transform domain method or a characteristic based method. For example: firstly, extracting the characteristics of two images to obtain characteristic points; finding matched characteristic point pairs by carrying out similarity measurement; then obtaining image space coordinate transformation parameters through the matched feature point pairs; and finally, carrying out image registration by the coordinate transformation parameters. The image registration can adopt a gray-scale information-based method, a transform domain method or a characteristic-based method.
In addition, the three-dimensional face registration module corresponds the face depth information to the two-dimensional face image according to the feature point corresponding relation between the two-dimensional face video picture and the two-dimensional face image after registration, so that the two-dimensional face image registered in the identity database is additionally recorded as the three-dimensional face image.
With the development of image acquisition technology, the acquisition of three-dimensional face data becomes more convenient, but the operation efficiency of a face recognition system can be reduced to a certain extent by the huge data amount contained in the three-dimensional face data. How to keep the geometric information of the face curved surface in the three-dimensional face data model as much as possible in the mapping process, and meanwhile, the use of a mature identification method in a two-dimensional space becomes a problem to be solved urgently at present. Therefore, in one embodiment, the three-dimensional face checking module comprises: the device comprises a dimension reduction mapping unit, an energy value extraction unit and a chi-square statistic unit.
The three-dimensional face video pictures collected by the video information collection equipment and the three-dimensional face images registered in the identity database are all three-dimensional face curved surface models, and the three-dimensional face model data comprises coordinate information of each point in the three-dimensional face curved surface in a three-dimensional space coordinate, information of a triangular patch formed by corresponding sequence points, coordinate information of a corresponding model two-dimensional texture mapping and texture points corresponding to patch information.
The dimensionality reduction mapping unit is used for conformally mapping a face model of the three-dimensional face image to a two-dimensional plane circular area to obtain a first face image based on a Richardy curvature stream, and conformally mapping the face model of the three-dimensional face video picture to the two-dimensional plane circular area to obtain a second face image, namely calculating a two-dimensional plane which is conformally equivalent to the original curved surface.
The Ricci curvature (Ricci curvature) is the sum of n-1 cross-sectional curvatures of an n-dimensional Riemann manifold. The essence of the richi curvature flow is the process of Riemann measurement deformation of each point of the curved surface, and the deformation degree is determined by the difference value of the Gaussian curvatures of each point of the curved surface, so that the Gaussian curvatures of each point of the curved surface are uniformly converted to each point of the curved surface along with the time under the control of a reaction-diffusion equation, and the curvature on the final curved surface is constant everywhere.
The geometric information of the three-dimensional face curved surface model comprises curvature information of each point in the model, so that the three-dimensional face model needs to be conformally mapped to two dimensions through the dimension reduction mapping unit, and then face recognition is carried out, so that the recognition accuracy and the recognition operation efficiency are improved.
Conformal geometry is essentially an intrinsic invariant between curved surfaces during transformation or mapping, the mapping process is mathematically equivalent to a smooth bijection from a certain neighborhood of a curved surface to a certain neighborhood of another curved surface or plane, and the parameterization process inevitably brings about a certain degree of distortion: angular distortion and area distortion. The mapping in which the angle is not distorted is called conformal mapping or conformal mapping, that is, the role played by the dimension reduction mapping unit.
Specifically, the existing dimension reduction algorithm inevitably causes geometric information of the face curved surface to be lost to a certain extent in the dimension reduction process, and therefore, in order to solve the problem, in an embodiment, the dimension reduction mapping unit includes: the device comprises a radius calculation subunit, an inversion distance calculation subunit, a curvature calculation subunit, a measurement length calculation subunit and a traversal subunit.
Under the condition that the human face posture is normal, the nose point of the human face curved surface model is the highest point of the whole model, so when a space rectangular coordinate system of the three-dimensional human face model is established, a plane where two line segments with the longest connecting line of two points in the human face curved surface are located is selected as a plane where an XOY plane is located, and the maximum Z value is extracted through analysis to obtain the nose point coordinate of the human face model. And setting the nose tip point as P, inputting the P point as a starting point into a Ricci curvature flow algorithm calculation process, and conformally mapping the three-dimensional face model to a two-dimensional plane disc.
Let there be a triangular mesh M ═ V, E, F, where V is the set of all vertices in the triangular mesh, E is the set of all edges in the triangular mesh, F is the set of all faces in the triangular mesh, V is the set of all faces in the triangular meshiIs the ith vertex in set V, eijIs connecting the vertices viAnd vjEdge of (f)ijkIs a vertex vi、vj、vkThe resulting facets.
And the radius calculation subunit is used for calculating an initial Riemann measurement radius according to the side length of the initial triangular mesh of the face curved surface and the vertex of each triangular mesh. Wherein, the Riemann measurement (circlepacking) radius calculation formula is as follows:
wherein lijAs edge e of a triangular meshijLength of side of likAnd ljkAnd so on. And the radius of the initial Riemann metric (i.e., vertex v) computed by the radius computation subunitiThe riemann metric radius of (d), the initial circular mode radius) is:
and the radius calculation subunit calculates to obtain the Riemann measurement corresponding to the point through the original triangulation measurement of the curved surface.
And the reverse distance calculation subunit is used for calculating the reverse distance of each grid edge according to the initial Riemannian measurement radius and the edge length of the corresponding triangular grid. The reverse distance I is a weight of the side length of the triangular mesh where the centers of the two circles are located in Riemann measurement. The inverse distance of the grid edge can be obtained by calculating through the cosine theorem, and the calculation formula is as follows:
during conformal mapping, only the size of the radius of the circular pattern is changed, but the reversal distance of each edge is kept unchanged.
And the curvature calculating subunit is used for calculating an internal angle of the triangular mesh according to the inversion distance and calculating the discrete Gaussian curvature of each vertex according to the internal angle. Under the discrete condition, the angle value of the vertex of each triangulation can be calculated by the cosine theorem of the side length, and the inner angle calculation formula is as follows:
wherein,
is a vertex vkFace f ofijkThe internal angle of (a).
The discrete gaussian curvature is a differential angle of a triangular patch adjacent to a vertex on the curved surface, and therefore, for a vertex in the curved surface, the differential angle is a difference between the sum of angles around the vertex and 2 pi, and for a boundary vertex, the differential angle is a difference between the sum of angles around the vertex and pi. The discrete gaussian curvature is calculated as:
wherein, KiRepresenting a vertex viThe curvature of the gaussian of (a) is,
is the boundary of a curved surface.
In the discrete case, the gaussian curvature of each point on the curved surface can be any value, but the total curvature of the curved surface remains unchanged during the variation, as shown in the following formula,
wherein A isijkIs a curved surface formed by a vertex vi、vj、vkAnd the area of the formed triangular patch, wherein lambda is a constant corresponding to the geometric background of the curved surface, lambda is 0 when the curved surface is in an Euclidean background, lambda is-1 when the curved surface is in a hyperbolic geometric background, and chi (M) is the Eulerian index of the curved surface. And the curvature calculation subunit calculates the Gaussian curvature of the point through the original triangulation measurement of the curved surface.
The measurement length calculation subunit is used for adjusting the radius of the circle according to the initial Gaussian curvature and the difference value between the target curvatures, further calculating the side length measured by the current triangular mesh according to the new radius of the circle, the radius of the circle of the adjacent point and the reversal distance of the side where the two circle centers are located, and calculating the angle of each vertex.
The vertex initial energy value is a logarithmic value of the radius of the initial circular mode defined by each point, and the calculation formula of the initial energy value is as follows:
ui＝logγi
changing the magnitude of the energy value of each point according to the magnitude of the initial Gaussian curvature and the difference value between the target curvatures, thereby changing the magnitude of the radius of the circle of each point as shown in the following formula:
u′i＝ui+(K′-K)
in order to map the surface commonality into a two-dimensional planar circular area, the target gaussian curvature K' is set to 0.
The side length of the current triangular mesh measurement can be reversely deduced through the radius of the new circular mode, the size of the original mode of the adjacent points and the reversal distance of the edge where the two points are located, namely the side length of the triangular mesh of each point when the energy of each point is minimum and the point is closest to the target curvature value, and the calculation formula is as follows:
the current vertex angle values can be calculated again by calculating the length of the new plane metric.
The measurement length calculation subunit changes Riemann measurement through the difference between the initial Gaussian curvature and the Gaussian curvature set by the user, and accordingly conformal change of the curved surface mesh measurement is achieved. At the moment, the energy value of each point on the curved surface reaches the optimum, namely, the plane metric induced by the Riemann metric obtained after calculation is conformal and equivalent to the original three-dimensional triangular mesh metric, and finally the side length and the angle value of the triangular mesh are recalculated according to the final Riemann metric to realize the embedding of the curved surface.
And the traversal subunit is used for controlling all the subunits to traverse all the vertexes on the face curved surface to obtain a plane measurement which is conformally equivalent to the triangular mesh measurement in the three-dimensional face model, and embedding the plane measurement into a two-dimensional plane circular area according to the triangular mesh corner relationship to obtain a face image subjected to conformal mapping.
The energy value extraction unit is used for respectively calculating the energy values of the vertexes of each triangular mesh in the planar models of the first human face image and the second human face image in conformal calculation, and further obtaining the energy minimum value of the multiple energy values of the vertexes of each triangular mesh.
After the conformal mapping is completed, the original three-dimensional face curved surface is mapped into a two-dimensional plane disc, and the Gaussian curvatures, namely the differential angles, of all points on the original three-dimensional face curved surface are converted into angle values of all vertexes of a triangle adjacent to the points through Ricci curvature sag. And constructing a Vertex Energy Minimum Pattern (VEMP) by extracting and mapping the Energy value of each Vertex in the rear plane model, thereby completing the feature extraction of the face model.
Specifically, in one embodiment, the energy value extraction unit encodes the calculated energy values of the vertices of the triangular mesh and assigns weights to the encoded energy values, calculates a plurality of initial vertex energy values of the vertices of the triangular mesh according to the weights and different encoding sequences, and uses a minimum value of the plurality of initial vertex energy values as an energy minimum value of the vertices of the triangular mesh.
In triangulation, each vertex may be adjoined by n triangular patches, so the number of vertices adjacent to the original vertex is 2 n. Energy value u after conformal mapping is completed by central vertex0(logarithmic form of Riemann measurement radius) as a threshold value, and constructing a vertex energy unit, wherein the energy values of adjacent 2n vertexes are respectively u1,u2…u2nIf the energy value of a vertex adjacent to the vertex is greater than or equal to the central vertex u0The energy value of (1) is coded as 1, and the energy value smaller than the center vertex is coded as 0. The calculation formula of the energy code is as follows:
T≈(S(u1-u0),S(u2-u0)…S(u2n-u0))
wherein, coding
After obtaining the energy value of the vertex adjacent to each vertex, all S (u) are generated into a binary vertex energy code, and then each code S (u) is distributed with a weight value of 2iThe binary code is converted to a decimal value, generating its initial Vertex Energy Value (VEP):
because there are many vertices adjacent to the triangular patch, different fixed point Energy codes are generated if different coding sequences are adopted for the Energy values extracted from each Vertex, so that binary codes generated by all the Energy values are calculated and the minimum value is taken as a final characteristic, namely, a Vertex Energy Minimum Pattern (VEMP), and the calculation formula is as follows:
VEMP＝min{VEPi|i＝1,2…,2n}
the chi-square statistic unit is used for generating an energy feature histogram according to the energy minimum value of each vertex of the first face image and the second face image, and judging the similarity of the first face image and the second face image by calculating and comparing chi-square statistic of the energy feature histogram, so as to obtain a comparison result.
After the conformal mapping is completed, and the energy values of all points of the curved surface of the human face are extracted by using the VEMP algorithm, the special energy characteristic histograms are calculated and generated, and the characteristic histograms generated by different models have obvious difference so as to identify whether the two human face images are the same or not. Wherein the horizontal axis of the histogram represents the magnitude of the energy value and the vertical axis represents the frequency of occurrence corresponding to the energy value.
Chi-square statistic is a measure of the difference between the distribution of data and a selected expected or assumed distribution, and can be used for calculating the similarity between a corresponding three-dimensional face image extracted from an identity database by a certificate reader and a three-dimensional face video image of a licensee collected by video information collection equipment, and when the similarity meets a set threshold, the comparison result between the two can be judged as follows: determining that the bearer is the owner of the document; and when the similarity fails to meet the set threshold, the comparison result of the two is judged as follows: it is determined that the bearer is not the owner of the document.
In the embodiment, aiming at the condition that the information of the three-dimensional face model is lost to different degrees in the dimension reduction process of the traditional three-dimensional face recognition algorithm, the three-dimensional face model is conformally mapped into the two-dimensional plane circular area by using a computation conformal geometric method based on Ricci curvature flow, and the operation mode can convert Gaussian curvature angle values of all points of the face curved surface into triangular mesh measurement adjacent to all points in the mapping process, so that the geometric information of the three-dimensional face curved surface is well reserved in the mapping process, and complete information resources are provided for the subsequent recognition process. And the deformation caused by the change of the dimensionality between the result of the conformal mapping completed by the Ricci curvature flow and the original three-dimensional face curved surface model is small, so that the geometric information in the three-dimensional face model is successfully converted into a two-dimensional plane disc. After conformal mapping is completed, extracting geometric data in each triangular patch in the model by setting an energy circular area, generating a feature histogram by counting the size change of each energy value on the whole face curved surface in the conformal mapping process of the whole face curved surface, completing feature extraction by generating the feature histogram, and finally completing the identification process of the face model to be matched (a three-dimensional face video picture) and the face model (a three-dimensional face image) in a standard face database by using chi-square statistics; in addition, even if the illumination intensities of the three-dimensional face video pictures are different, the generated energy feature histograms have no obvious difference, so that the system can correctly identify the face without being influenced by the illumination intensities, and the robustness of the system to the illumination intensities is improved.
An embodiment of the testimonial-to-testimonial inspection method disclosed in the present application is described in detail below with reference to fig. 2. This embodiment is used to implement the aforementioned embodiment of the testimonial verification system. As shown in fig. 2, the method disclosed in this embodiment includes the following steps:
step 100, reading identity information contained in a certificate held by a certificate holder, identifying, and extracting a pre-registered two-dimensional face image from an identity database according to an identification result;
step 200, collecting a three-dimensional face video picture of a licensee in real time, wherein the three-dimensional face video picture comprises a two-dimensional face video picture and corresponding face depth information;
step 400, under the condition that the two-dimensional image comparison result shows that the two-dimensional image comparison result meets the matching condition, generating a three-dimensional face image based on a pre-registered two-dimensional face image and corresponding face depth information and registering the three-dimensional face image;
In one embodiment, before comparing the acquired two-dimensional face video picture with the pre-registered two-dimensional face image in step 300, a face picture with the largest face frame area is extracted from the acquired two-dimensional face video picture as a comparison target image.
In one embodiment, the comparing 300 the captured two-dimensional face video frame with the pre-registered two-dimensional face image includes: extracting characteristic points with face representativeness from the collected two-dimensional face video picture and the pre-registered two-dimensional face image respectively, and registering the two-dimensional face video picture and the two-dimensional face image through matching of the characteristic points;
in step 400, generating and registering a three-dimensional face image based on a pre-registered two-dimensional face image and corresponding face depth information includes: and according to the feature point correspondence between the registered two-dimensional face video picture and the two-dimensional face image, corresponding the face depth information to the two-dimensional face image, thereby additionally recording the two-dimensional face image registered in the identity database into a three-dimensional face image.
In one embodiment, the step 500 of comparing the three-dimensional face image with the three-dimensional face video picture of the witness collected by the video information collection device during the inspection comprises:
based on the Rich curvature stream, conformally mapping a face model of the three-dimensional face image to a two-dimensional plane circular area to obtain a first face image, and conformally mapping the face model of the three-dimensional face video picture to the two-dimensional plane circular area to obtain a second face image;
respectively calculating energy values of vertexes of each triangular mesh in the plane models of the first face image and the second face image in conformal calculation, and further obtaining an energy minimum value in a plurality of energy values of the vertexes of each triangular mesh;
and generating an energy feature histogram according to the energy minimum value of each vertex of the first face image and the second face image, and judging the similarity of the first face image and the second face image by calculating and comparing chi-square statistics of the energy feature histograms so as to obtain a comparison result.
In one embodiment, the step 500 of conformally mapping the face model of the face image to the two-dimensional planar circular region based on the reed curvature stream comprises:
calculating an initial Riemann measurement radius according to the side length of the initial triangular mesh of the face curved surface and the vertex of each triangular mesh;
calculating the reversal distance of each grid edge according to the initial Riemann measurement radius and the edge length of the corresponding triangular grid;
calculating an inner angle of the triangular mesh according to the inversion distance, and calculating the discrete Gaussian curvature of each vertex according to the inner angle;
adjusting the radius of the circle according to the difference between the initial Gaussian curvature and the target curvature, further calculating the side length measured by the current triangular mesh according to the new radius of the circle, the radius of the circle of the adjacent point and the reversal distance of the side where the two circle centers are located, and calculating the angle of each vertex;
and controlling all the subunits to traverse all vertexes on the face curved surface to obtain plane measurement which is conformally equivalent to triangular mesh measurement in the three-dimensional face model, and embedding the plane measurement into a two-dimensional plane circular area according to the corner relationship of the triangular mesh to obtain a face image subjected to conformal mapping.
In one embodiment, the step 500 of obtaining an energy minimum value of the plurality of energy values of each triangle mesh vertex comprises:
and coding the calculated energy values of the triangular mesh vertexes, respectively distributing weights, calculating a plurality of initial vertex energy values of the triangular mesh vertexes according to the weights and different coding sequences, and taking the minimum value of the plurality of initial vertex energy values as the energy minimum value of the triangular mesh vertexes.
In this document, "first", "second", and the like are used only for distinguishing one from another, and do not indicate their degree of importance, order, and the like.
The division of the modules and units herein is only one division of logical functions, and other divisions may be possible in actual implementation, for example, a plurality of modules and/or units may be combined or integrated in another system. The modules and units described as separate parts may be physically separated or not. The components displayed as cells may or may not be physical cells, and may be located in a specific place or distributed in grid cells. Therefore, some or all of the units can be selected according to actual needs to implement the scheme of the embodiment.
The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present application should be covered within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (8)
1. The utility model provides a testimony of a witness unification inspection system based on real-time face identification which characterized in that includes:
the certificate reader is used for reading identity information contained in a certificate held by a certificate holder, identifying the identity information and extracting a pre-registered two-dimensional face image from an identity database according to an identification result;
the video information acquisition equipment is used for acquiring a three-dimensional face video picture of a licensee in real time, wherein the three-dimensional face video picture comprises a two-dimensional face video picture and corresponding face depth information;
the two-dimensional image comparison module is used for comparing the acquired two-dimensional face video picture with the pre-registered two-dimensional face image;
the three-dimensional face registration module is used for generating a three-dimensional face image based on the pre-registered two-dimensional face image and the corresponding face depth information and registering the three-dimensional face image and the pre-registered two-dimensional face image under the condition that the two-dimensional image comparison result shows that the two-dimensional image and the pre-registered two-dimensional face image meet the matching condition;
the three-dimensional face checking module is used for extracting a three-dimensional face image registered by the three-dimensional face registration module from the identity database according to a recognition result obtained by recognizing the read identity information by the certificate reader when checking the certificate holder, comparing the three-dimensional face image with a three-dimensional face video image of the certificate holder acquired by the video information acquisition equipment during checking, and judging whether the certificate holder is a certificate owner according to a three-dimensional image comparison result;
wherein, the three-dimensional face checking module comprises: the dimensionality reduction mapping unit is used for conformally mapping the face model of the three-dimensional face image to a two-dimensional plane circular area to obtain a first face image based on a Rich curvature stream, and conformally mapping the face model of the three-dimensional face video picture to the two-dimensional plane circular area to obtain a second face image; an energy value extraction unit, configured to calculate an energy value generated by each triangular mesh vertex in the planar model of the first face image and the second face image in conformal calculation, respectively, and further obtain an energy minimum value among a plurality of energy values of each triangular mesh vertex; and the chi-square statistic unit is used for generating an energy feature histogram according to the energy minimum value of each vertex of the first face image and the second face image, and judging the similarity between the first face image and the second face image by calculating and comparing chi-square statistic of the energy feature histogram so as to obtain the comparison result.
2. The system according to claim 1, wherein the two-dimensional image comparison module extracts feature points having face representativeness from the acquired two-dimensional face video picture and the pre-registered two-dimensional face image, respectively, and performs registration between the two-dimensional face video picture and the two-dimensional face image through matching of the feature points; and,
and the three-dimensional face registration module corresponds the face depth information to the two-dimensional face image according to the feature point corresponding relation between the two-dimensional face video picture and the two-dimensional face image after registration, so that the two-dimensional face image registered in the identity database is recorded as the three-dimensional face image.
3. The system of claim 1, wherein the dimension reduction mapping unit comprises:
the radius calculation subunit is used for calculating an initial Riemann measurement radius according to the side length of an initial triangular mesh of the face curved surface and the vertex of each triangular mesh;
the reverse distance calculation subunit is used for calculating the reverse distance of each grid edge according to the initial Riemannian measurement radius and the edge length of the corresponding triangular grid;
the curvature calculation subunit is used for calculating an inner angle of the triangular grid according to the inversion distance and calculating the discrete Gaussian curvature of each vertex according to the inner angle;
the measurement length calculation subunit is used for adjusting the radius of the circle according to the initial discrete Gaussian curvature and the difference value between the target curvatures, further calculating the side length measured by the current triangular grid according to the new radius of the circle, the radius of the circle of the adjacent point and the reversal distance of the side where the two circle centers are located, and calculating the angle of each vertex;
and the traversal subunit is used for controlling all the subunits to traverse all the vertexes on the curved surface of the human face to obtain a planar measurement which is conformally equivalent to the triangular mesh measurement in the three-dimensional human face model, and embedding the planar measurement into a two-dimensional planar circular area according to the triangular mesh corner relationship to obtain a human face image after conformal mapping.
4. The system according to claim 1 or 3, wherein the energy value extracting unit encodes the calculated energy values of the vertices of the triangular mesh and assigns weights to the encoded energy values, calculates a plurality of initial vertex energy values of the vertices of the triangular mesh based on the weights and different encoding orders, and takes a minimum value of the plurality of initial vertex energy values as an energy minimum value of the vertices of the triangular mesh.
5. A testimony of a witness unification inspection method based on real-time face identification, its characterized in that includes:
reading identity information contained in a certificate held by a certificate holder, identifying, and extracting a pre-registered two-dimensional face image from an identity database according to an identification result;
acquiring a three-dimensional face video picture of a licensee in real time, wherein the three-dimensional face video picture comprises a two-dimensional face video picture and corresponding face depth information;
comparing the collected two-dimensional face video picture with the pre-registered two-dimensional face image;
under the condition that the two-dimensional image comparison result shows that the two images accord with the matching condition, generating a three-dimensional face image based on the pre-registered two-dimensional face image and the corresponding face depth information and registering;
when a certificate holder is checked, extracting the three-dimensional face image registered in the previous step from the identity database according to the recognition result obtained by recognizing the read identity information by the certificate reader, comparing the three-dimensional face image with the three-dimensional face video image of the certificate holder collected during checking, and judging whether the certificate holder is a certificate owner according to the comparison result of the three-dimensional image;
wherein, the comparing the three-dimensional face image with the three-dimensional face video picture of the witness collected during the checking comprises:
based on a Rich curvature stream, conformally mapping a face model of the three-dimensional face image to a two-dimensional plane circular area to obtain a first face image, and conformally mapping the face model of the three-dimensional face video picture to the two-dimensional plane circular area to obtain a second face image;
respectively calculating energy values of vertexes of each triangular mesh in the plane models of the first face image and the second face image in conformal calculation, and further obtaining an energy minimum value in a plurality of energy values of the vertexes of each triangular mesh;
generating an energy feature histogram according to the energy minimum value of each vertex of the first face image and the second face image, and judging the similarity of the first face image and the second face image by calculating and comparing chi-square statistics of the energy feature histogram, so as to obtain the comparison result.
6. The method of claim 5, wherein said comparing said captured two-dimensional face video frame to said pre-registered two-dimensional face image comprises: extracting characteristic points with face representativeness from the collected two-dimensional face video picture and the pre-registered two-dimensional face image respectively, and registering the two-dimensional face video picture and the two-dimensional face image through matching of the characteristic points;
and the generating and registering a three-dimensional face image based on the pre-registered two-dimensional face image and the corresponding face depth information comprises: and according to the feature point correspondence between the two-dimensional face video picture and the two-dimensional face image after registration, corresponding the face depth information to the two-dimensional face image, thereby registering the two-dimensional face image registered in an identity database as the three-dimensional face image.
7. The method of claim 5, wherein conformally mapping a face model of a face image to a two-dimensional planar circular region based on a stream of richness curvatures comprises:
calculating an initial Riemann measurement radius according to the side length of an initial triangular mesh of the face curved surface and the vertex of each triangular mesh;
calculating the reversal distance of each grid edge according to the initial Riemannian measurement radius and the edge length of the corresponding triangular grid;
calculating an inner angle of the triangular mesh according to the inversion distance, and calculating a discrete Gaussian curvature of each vertex according to the inner angle;
adjusting the radius of the circle according to the initial discrete Gaussian curvature and the difference value between the target curvatures, further calculating the side length measured by the current triangular mesh according to the new radius of the circle, the radius of the circle of the adjacent point and the reversal distance of the side where the two circle centers are located, and calculating the angle of each vertex;
traversing all vertexes on the face curved surface to obtain plane measurement which is conformal equivalent to triangular mesh measurement in the three-dimensional face model, and embedding the plane measurement into a two-dimensional plane circular area according to the triangular mesh corner relation to obtain a face image subjected to conformal mapping.
8. The method of claim 5 or 7, wherein obtaining an energy minimum of the plurality of energy values for each triangular mesh vertex comprises:
and coding the calculated energy values of the triangular mesh vertexes, respectively distributing weights, calculating a plurality of initial vertex energy values of the triangular mesh vertexes according to the weights and different coding sequences, and taking the minimum value of the plurality of initial vertex energy values as the energy minimum value of the triangular mesh vertexes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910944252.2A CN110866442B (en) | 2019-09-30 | 2019-09-30 | Real-time face recognition-based testimony-of-person integrated checking system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910944252.2A CN110866442B (en) | 2019-09-30 | 2019-09-30 | Real-time face recognition-based testimony-of-person integrated checking system and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110866442A CN110866442A (en) | 2020-03-06 |
| CN110866442B true CN110866442B (en) | 2021-06-04 |
Family
ID=69652687
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910944252.2A Active CN110866442B (en) | 2019-09-30 | 2019-09-30 | Real-time face recognition-based testimony-of-person integrated checking system and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110866442B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111329442A (en) * | 2020-03-16 | 2020-06-26 | 广东小天才科技有限公司 | Eyesight health detection method and device and electronic equipment |
| CN112580436B (en) * | 2020-11-25 | 2022-05-03 | 重庆邮电大学 | Electroencephalogram signal domain adaptation method based on Riemann manifold coordinate alignment |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108764140A (en) * | 2018-05-29 | 2018-11-06 | 中北大学 | A kind of three-dimensional face identification method based on Ricci curvature flows |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN206515838U (en) * | 2016-09-07 | 2017-09-22 | 上海微桥电子科技有限公司 | A kind of testimony of a witness unifies instrument |
| CN108052814A (en) * | 2017-12-07 | 2018-05-18 | 四川大学 | A kind of 3D authentication systems |
| CN109670487A (en) * | 2019-01-30 | 2019-04-23 | 汉王科技股份有限公司 | A kind of face identification method, device and electronic equipment |
| CN110020601B (en) * | 2019-03-06 | 2023-08-22 | 创新先进技术有限公司 | Three-dimensional face bottom-remaining registration method, device, terminal and server |
-
2019
- 2019-09-30 CN CN201910944252.2A patent/CN110866442B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108764140A (en) * | 2018-05-29 | 2018-11-06 | 中北大学 | A kind of three-dimensional face identification method based on Ricci curvature flows |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110866442A (en) | 2020-03-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9141844B2 (en) | System and method for three-dimensional biometric data feature detection and recognition | |
| Berretti et al. | Matching 3D face scans using interest points and local histogram descriptors | |
| Dagnes et al. | Occlusion detection and restoration techniques for 3D face recognition: a literature review | |
| AU2004271639B2 (en) | Systems and methods for biometric identification using handwriting recognition | |
| CN111126240B (en) | Three-channel feature fusion face recognition method | |
| Islam et al. | A review of recent advances in 3D ear-and expression-invariant face biometrics | |
| Zhou et al. | An efficient 3-D ear recognition system employing local and holistic features | |
| CN107909031B (en) | Crime scene fingerprint line leaving area frequency dynamic reconstruction method | |
| Diwakar et al. | An extraction and recognition of tongue-print images for biometrics authentication system | |
| Bhanu et al. | Human ear recognition by computer | |
| CN102542243A (en) | LBP (Local Binary Pattern) image and block encoding-based iris feature extracting method | |
| Krishneswari et al. | A review on palm print verification system | |
| CN110866442B (en) | Real-time face recognition-based testimony-of-person integrated checking system and method | |
| Bai et al. | Person recognition using 3-D palmprint data based on full-field sinusoidal fringe projection | |
| CN110222660B (en) | Signature authentication method and system based on dynamic and static feature fusion | |
| Zhang et al. | Efficient 3D dental identification via signed feature histogram and learning keypoint detection | |
| CN106778491B (en) | The acquisition methods and equipment of face 3D characteristic information | |
| CN109598235B (en) | Finger vein image authentication method and device | |
| JP2015162012A (en) | Face matching device, face matching method and program | |
| Zhang et al. | Deep dense multi-level feature for partial high-resolution fingerprint matching | |
| Szczepanik et al. | Data management for fingerprint recognition algorithm based on characteristic points’ groups | |
| Jain | Automatic Fingerprint Matching Using Extended Feature Set | |
| Radouane et al. | Fusion of Gabor filter and steerable pyramid to improve iris recognition system | |
| Yan et al. | A novel bimodal identification approach based on hand-print | |
| Dua'a Hamed et al. | Finger knuckle recognition, a review on prospects and challenges based on PolyU dataset |
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 |