CN105447466B - A kind of identity integrated recognition method based on Kinect sensor - Google Patents
A kind of identity integrated recognition method based on Kinect sensor Download PDFInfo
- Publication number
- CN105447466B CN105447466B CN201510862672.8A CN201510862672A CN105447466B CN 105447466 B CN105447466 B CN 105447466B CN 201510862672 A CN201510862672 A CN 201510862672A CN 105447466 B CN105447466 B CN 105447466B
- Authority
- CN
- China
- Prior art keywords
- accredited personnel
- information
- personnel
- indicate
- human
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
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
-
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of identity integrated recognition method based on Kinect sensor, the following steps are included: one, using Kinect sensor obtain accredited personnel multiple groups characteristics of human body, every group of characteristics of human body includes human face image information, the colour of skin/color development information and Human Height information;Two, Haar-Like feature is extracted based on human face image information and obtains the recognition of face classifier result of accredited personnel by SVM algorithm;Three, it is based on the colour of skin/color development information, obtains the colour of skin/color development mixed Gauss model of accredited personnel;Four, it is based on Human Height information, obtains the height average and standard deviation of accredited personnel;It five, will be Step 2: the result of step 3 and step 4 be stored in database to complete the information registering of accredited personnel;Six, the characteristics of human body of Kinect sensor capture current persons, the identity for determining current persons that the characteristics of human body of current persons is made comparisons with the registration information of accredited personnel in database are utilized.
Description
Technical field
The present invention relates to a kind of personal identification methods, are especially one kind based on Kinect sensor and merge height letter
The identity integrated recognition method of breath, the colour of skin/color development information and human face image information.
Background technique
In recent years, with the breakthrough development of the technologies such as computer, internet, artificial intelligence, robot is just gradually entered into
The normal sphere of life of the mesh of people, provides various types of services.Service humanoid robot application in, allow machine person to person into
The exchange of the various information of row be realize the premise of good service, and allow robot quickly, the identity of correct identification people, as owner,
Client, stranger etc. are the basic guarantees that machine person to person is exchanged to realize that differentiation is treated.
In artificial intelligence field, authentication can be realized by multiple technologies, such as input password, brush ID card, finger print identifying,
Iris recognition etc..Although these technologies are widely used, and uniqueness, confidentiality are preferable, but the service for serving people
For humanoid robot, and it is not suitable for.Reason is very simple, is exactly that people prefer to that nature can be carried out with robot as people
Exchange, rather than by by input password, swipe the card etc. it is cumbersome in a manner of obtain manipulation power.It can be seen that above-mentioned identification authentication mode is only
The means for obtaining control as robot administrator or maintenance personnel are more appropriate.
It is mainly based upon the recognition of face of image using more authentication means in robot field at present, has
Simply, natural, non-contacting advantage.But there is inherent shortcomings for recognition of face, on the one hand, identification process needs identified person
Cooperation, to provide front face, and recognition effect is poor in the bad situation of light, and accuracy rate is lower;On the other hand, machine
Device people is easy to be cheated by invader using photo.Especially under the application of some household scenes, the service object of robot is main
For kinsfolk, kinsfolk only is identified by recognition of face means, it is avoided that not needing owner is frequently being required to provide just
Face cooperation, use is very inconvenient, and flexibility is poor, and experience comfort level is lower.
Summary of the invention
The object of the present invention is to provide a kind of identity integrated recognition method based on Kinect sensor has and realizes and hold
Easily, the advantage that recognition speed is fast, accuracy rate is high is not needed using service robot of the invention under family's application scenarios
Continually owner is required to cooperate, can be achieved with identity detection identification in most cases, practicability is stronger.
Be not suitable for service humanoid robot use to solve authentication mode commonly used in the prior art, and it is single based on figure
The face recognition technology of picture needs owner frequently to cooperate, and in the case where light is bad, recognition effect is poor, and is easy to be invaded
The technical issues of person is cheated using photo, a kind of identity integrated recognition method based on Kinect sensor provided by the invention,
Including registration process and identification process, and specifically includes the following steps:
One, it allows accredited personnel's multi-angle rotation face before Kinect sensor, and does different limbs in different location
Movement, to obtain the multiple groups characteristics of human body of the accredited personnel, every group of characteristics of human body includes human face image information, the colour of skin/color development
Information and Human Height information;
Two, the multiple groups human face image information based on the accredited personnel extracts Haar-Like feature and passes through SVM algorithm list
Solely training face recognition classifier, to obtain the recognition of face classifier result of the accredited personnel;
Three, the multiple groups colour of skin/color development information based on the accredited personnel passes through the accumulative colour of skin/hair for obtaining the accredited personnel
Mixture of colours Gauss model;
Four, the multiple groups Human Height information based on the accredited personnel, the height for obtaining the accredited personnel by calculating are average
Value and standard deviation;
It five, will be Step 2: the result that step 3 and step 4 obtain be stored in database to complete the information of the accredited personnel
It registers, and completes the information registering of all accredited personnel according to the logon mode of the accredited personnel;
Six, after the completion of registering, using the characteristics of human body of Kinect sensor capture current persons, by the human body of current persons
Feature is made comparisons with the registration information of accredited personnel in database, and the identity of current persons is determined according to comparison result.
Further, a kind of identity integrated recognition method based on Kinect sensor of the present invention, wherein in step 1
In, the human face image information for obtaining accredited personnel is realized by mode in detail below:
(1) using depth image and color image of the Kinect sensor acquisition comprising accredited personnel, and according to depth map
The human skeleton artis information of depth data reduction accredited personnel as in, wherein
Torso portion includes the crown, lower jaw, chest, abdomen, hip, is successively indicated with C1, C2, C3, C4, C5;
Left-hand part includes left hand finger tip, left finesse, left elbow joint, left shoulder joint, is successively indicated with L1, L2, L3, L4;
Right hand portion includes right hand finger tip, right finesse, right elbow joint, right shoulder joint, is successively indicated with R1, R2, R3, R4;
Left leg section includes left foot point, left foot wrist, left knee joint, left hip joint, is successively indicated with E1, E2, E3, E4;
Right leg section includes right crus of diaphragm point, right crus of diaphragm wrist, right knee joint, right hip joint, is successively indicated with F1, F2, F3, F4;
(2) using the line of two artis of C1 in the human skeleton of accredited personnel and C2 as axis, using human body segmentation side
Method extracts the human body head region in color image, as human body head image;
(3) judge whether human body head image includes face using face recognition algorithms, grab people if including face
Face image, as the human face image information of accredited personnel, otherwise it is assumed that not including face.
Further, a kind of identity integrated recognition method based on Kinect sensor of the present invention, wherein in step 1
In, the colour of skin/color development information for obtaining accredited personnel is realized by mode in detail below:
(1) the human body head image of accredited personnel is converted into YCbCr colour gamut from RGB color domain, and is directed to human body head figure
Each pixel as in judges whether its CbCr chrominance component belongs to basic skin distribution U (Cb, Cr), marks if belonging to
It is 1,0 is labeled as if being not belonging to;
(2) according to the judgement of step (1) and label as a result, using it is all mark be pixel as one gather, and
The mean value and the corresponding covariance matrix of CbCr for calculating CbCr chrominance component, as colour of skin list Gauss model, wherein CbCr is colored
Component mean value is usedIt indicates, covariance matrix σ1 2It indicates, colour of skin list Gauss model N1(μ1, σ1 2) table
Show;
(3) according to the judgement of step (1) and label as a result, using it is all mark be pixel as one gather, and
The mean value and the corresponding covariance matrix of CbCr for calculating CbCr chrominance component, as color development list Gauss model, wherein CbCr is colored
Component mean value is usedIt indicates, covariance matrix variance σ2 2It indicates, color development list Gauss model N2(μ2,
σ2 2) indicate.
Further, a kind of identity integrated recognition method based on Kinect sensor of the present invention, wherein in step 3
In, the colour of skin/color development mixed Gauss model is N=(μ1, σ1 2, μ2, σ2 2)。
Further, a kind of identity integrated recognition method based on Kinect sensor of the present invention, wherein in step 1
In, the Human Height information for obtaining accredited personnel is realized by mode in detail below:
(1) artis in human skeleton is divided into five groups, the 1st group is (C1, C2, C3, C4, C5), the 2nd group for (L1,
L2, L3, L4), the 3rd group is (R1, R2, R3, R4), and the 4th group is (E1, E2, E3, E4), and the 5th group is (F1, F2, F3, F4);
(2) least square method fitting three-dimensional space straight line is respectively adopted to each group joint point set, and calculated respective straight
Line error of fitting is denoted as Δ 1, Δ 2, Δ 3, Δ 4, Δ 5 respectively;
(3) when all error deltas 1, Δ 2, Δ 3, Δ 4, Δ 5 are respectively less than given threshold T, then it is assumed that human body is in each pass
The straight configuration of section, and the Human Height indicated with H is calculated as follows;
α=(Δ4+Δ5)/(Δ1+Δ2+Δ3+Δ4+Δ5) (6)
H=α (H1+max(H2, H3))+2(1-α)max(A1+A2) (7)
In above-mentioned formula (1) into (5),Indicate the three-dimensional space distance between two joint point C1 and C2;Table
Show the three-dimensional space distance between two joint point C2 and C3;Indicate the three-dimensional space distance between two joint point C3 and C4;Indicate the three-dimensional space distance between two joint point C4 and C5;Indicate the three-dimensional space between two joint point E1 and E2
Between distance;Indicate the three-dimensional space distance between two joint point E2 and E3;It indicates between two joint point E3 and E4
Three-dimensional space distance;Indicate the three-dimensional space distance between two joint point F1 and F2;Indicate two joint point F2 and
Three-dimensional space distance between F3;Indicate the three-dimensional space distance between two joint point F3 and F4;Indicate two joint
Three-dimensional space distance between point L1 and L2;Indicate the three-dimensional space distance between two joint point L2 and L3;It indicates
Three-dimensional space distance between two joint point L3 and L4;Indicate the three-dimensional space distance between two joint point L4 and C3;Indicate the three-dimensional space distance between two joint point R1 and R2;Indicate the three-dimensional space between two joint point R2 and R3
Between distance;Indicate the three-dimensional space distance between two joint point R3 and R4;It indicates between two joint point R4 and C3
Three-dimensional space distance.
Further, a kind of identity integrated recognition method based on Kinect sensor of the present invention, wherein in step 6
In, the characteristics of human body using Kinect sensor capture current persons, by accredited personnel in the people's body characteristics and database
Registration information make comparisons to determine the identity of current persons, specifically includes the following steps:
(1) the Human Height information and the colour of skin/color development information of current persons are obtained using Kinect sensor;
(2) pass through the height for calculating and obtaining current persons according to the Human Height information of current persons, inquire database
The registration information of middle accredited personnel simultaneously traverses corresponding [h-3 Δ h, the h+3 Δ h] range of each accredited personnel, judges whether to deposit
In the accredited personnel to match with current persons' height, if there is and there is uniqueness to be then directly identified as current persons pair
The accredited personnel answered terminates to identify and exports result;Following third step is then carried out if there is but without uniqueness;If no
There are matched accredited personnel then to carry out following 4th step;Wherein h indicates that the height average value of accredited personnel, Δ h indicate standard
Difference;
(3) according to the colour of skin of current persons/color development information, the colour of skin/color development mixed Gauss model of current persons is obtained,
According to there is the accredited personnel to match with current persons' height but not unique condition on the basis of step (2), determining and waiting
Select accredited personnel's range and judge whether there is with the colour of skin of current persons/hair color model unique match accredited personnel, if
Current persons are then identified as the accredited personnel by the matched accredited personnel of existence anduniquess, are terminated to identify and are exported result;If no
The matched accredited personnel of existence anduniquess then following 4th step of progress;
(4) phonetic order is issued, it is desirable that front face towards Kinect sensor, and is obtained and work as forefathers by current persons
The human face image information of member;
(5) according to the human face image information of current persons, the accredited personnel's information inquired in database simultaneously judges whether to deposit
In matched accredited personnel, and if so, current persons are identified as corresponding accredited personnel, terminate to identify and export as a result,
Current persons are then identified as strange personnel or current persons are required to re-register by accredited personnel if there is no match;
Wherein, Human Height information, the colour of skin/color development information and the face figure of current persons are obtained using Kinect sensor
As the implementation of information is identical as implementation when registration.
A kind of identity integrated recognition method based on Kinect sensor of the present invention compared with prior art, has following excellent
Point: identity integrated recognition method provided by the invention is based on Microsoft's Kinect sensor, and Kinect is using actively red
Outer line technology carries out depth finding, can effectively avoid the influence of illumination condition and shelter, can obtain in photographed scene in real time
Color image and depth image, the color image and depth image that the present invention is shot according to Kinect sensor extract human body
Human face image information, the colour of skin/color development information and height information, and in registration and identification process, melt in conjunction with much information breath
Close judgement, can effectively promote the accuracy rate of identification, have implement it is easy, unsophisticated, non-contact, without wearing the excellent of foreign object
Point, especially under family's application scenarios, using the robot of the method for the present invention without continually owner being required to cooperate, it will be able to
More convenient quickly to identify different kinsfolks, practicability is stronger.
Illustrated embodiment is to a kind of identity integrated recognition method based on Kinect sensor of the present invention with reference to the accompanying drawing
It is described in further detail:
Detailed description of the invention
Fig. 1 is a kind of human skeleton schematic diagram of the identity integrated recognition method based on Kinect sensor of the present invention;
Fig. 2 is a kind of identification flow chart of the identity integrated recognition method based on Kinect sensor of the present invention.
Specific embodiment
Below to carry the application that the robot of Kinect sensor identifies different home member under home environment, as
A kind of specific example mode of the identity integrated recognition method based on Kinect sensor of the present invention is specifically described.It needs first
It is noted that the Kinect sensor of Microsoft can sampling depth image and color image in real time, and by Kinect from
The SDK of band can effectively obtain its human body target within the vision and identify skeleton.
A kind of identity integrated recognition method based on Kinect sensor of the present invention, including register and identify two processes,
And specifically includes the following steps:
One, it allows accredited personnel's multi-angle rotation face before Kinect sensor, and does different limbs in different location
Movement, to obtain the multiple groups characteristics of human body of the accredited personnel, every group of characteristics of human body includes human face image information, the colour of skin/color development
Information and Human Height information;
Two, the multiple groups human face image information based on the accredited personnel extracts Haar-Like feature and passes through SVM algorithm list
Solely training face recognition classifier, to obtain the recognition of face classifier result of the accredited personnel;
Three, the multiple groups colour of skin/color development information based on the accredited personnel passes through the accumulative colour of skin/hair for obtaining the accredited personnel
Mixture of colours Gauss model;
Four, the multiple groups Human Height information based on the accredited personnel, the height for obtaining the accredited personnel by calculating are average
Value and standard deviation;
It five, will be Step 2: the result that step 3 and step 4 obtain be stored in database to complete the information of the accredited personnel
It registers, and completes the information registering of all accredited personnel according to the logon mode of the accredited personnel;
Six, after the completion of registering, using the characteristics of human body of Kinect sensor capture current persons, by the human body of current persons
Feature is made comparisons with the registration information of accredited personnel in database, and the identity of current persons is determined according to comparison result.
In the above step 1, the human face image information for obtaining accredited personnel, is realized by mode in detail below:
(1) using depth image and color image of the Kinect sensor acquisition comprising accredited personnel, and according to depth map
The human skeleton artis information of depth data reduction accredited personnel as in, as shown in Figure 1, wherein
Torso portion includes the crown, lower jaw, chest, abdomen, hip, is successively indicated with C1, C2, C3, C4, C5;
Left-hand part includes left hand finger tip, left finesse, left elbow joint, left shoulder joint, is successively indicated with L1, L2, L3, L4;
Right hand portion includes right hand finger tip, right finesse, right elbow joint, right shoulder joint, is successively indicated with R1, R2, R3, R4;
Left leg section includes left foot point, left foot wrist, left knee joint, left hip joint, is successively indicated with E1, E2, E3, E4;
Right leg section includes right crus of diaphragm point, right crus of diaphragm wrist, right knee joint, right hip joint, is successively indicated with F1, F2, F3, F4;
(2) using the line of two artis of C1 in the human skeleton of accredited personnel and C2 as axis, using human body segmentation side
Method extracts the human body head region in color image, as human body head image;
(3) judge whether human body head image includes face using face recognition algorithms, grab people if including face
Face image, as the human face image information of accredited personnel, otherwise it is assumed that not including face.
In the above step 1, the colour of skin/color development information for obtaining accredited personnel, is realized by mode in detail below:
(1) the human body head image of accredited personnel is converted into YCbCr colour gamut from RGB color domain, and is directed to human body head figure
Each pixel as in judges whether its CbCr chrominance component belongs to basic skin distribution U (Cb, Cr), marks if belonging to
It is 1,0 is labeled as if being not belonging to;
(2) according to the judgement of step (1) and label as a result, using it is all mark be pixel as one gather, and
The mean value and the corresponding covariance matrix of CbCr for calculating CbCr chrominance component, as colour of skin list Gauss model, wherein CbCr is colored
Component mean value is usedIt indicates, covariance matrix σ1 2It indicates, colour of skin list Gauss model N1(μ1, σ1 2) table
Show;
(3) according to the judgement of step (1) and label as a result, using it is all mark be pixel as one gather, and
The mean value and the corresponding covariance matrix of CbCr for calculating CbCr chrominance component, as color development list Gauss model, wherein CbCr is colored
Component mean value is usedIt indicates, covariance matrix σ2 2It indicates, color development list Gauss model N2(μ2, σ2 2) table
Show.
In above-mentioned steps three, the colour of skin/color development mixed Gauss model is N=(μ1, σ1 2, μ2, σ2 2)。
In the above step 1, the Human Height information for obtaining accredited personnel, is realized by mode in detail below:
(1) artis in human skeleton is divided into five groups, the 1st group is (C1, C2, C3, C4, C5), the 2nd group for (L1,
L2, L3, L4), the 3rd group is (R1, R2, R3, R4), and the 4th group is (E1, E2, E3, E4), and the 5th group is (F1, F2, F3, F4);
(2) least square method fitting three-dimensional space straight line is respectively adopted to each group joint point set, and calculated respective straight
Line error of fitting is denoted as Δ 1, Δ 2, Δ 3, Δ 4, Δ 5 respectively;
(3) when all error deltas 1, Δ 2, Δ 3, Δ 4, Δ 5 are respectively less than given threshold T, then it is assumed that human body is in each pass
The straight configuration of section, and the Human Height indicated with H is calculated as follows;
α=(Δ4+Δ5)/(Δ1+Δ2+Δ3+Δ4+Δ5) (6)
H=α (H1+max(H2, H3))+2(1-α)max(A1+A2) (7)
In formula (1) into (5),Indicate the three-dimensional space distance between two joint point C1 and C2;Indicate two
Three-dimensional space distance between artis C2 and C3;Indicate the three-dimensional space distance between two joint point C3 and C4;
Indicate the three-dimensional space distance between two joint point C4 and C5;Indicate two joint point E1 and E2 between three-dimensional space away from
From;Indicate the three-dimensional space distance between two joint point E2 and E3;Indicate the three-dimensional between two joint point E3 and E4
Space length;Indicate the three-dimensional space distance between two joint point F1 and F2;It indicates between two joint point F2 and F3
Three-dimensional space distance;Indicate the three-dimensional space distance between two joint point F3 and F4;Indicate two joint point L1 and
Three-dimensional space distance between L2;Indicate the three-dimensional space distance between two joint point L2 and L3;Indicate two joint
Three-dimensional space distance between point L3 and L4;Indicate the three-dimensional space distance between two joint point L4 and C3;Table
Show the three-dimensional space distance between two joint point R1 and R2;Indicate the three-dimensional space distance between two joint point R2 and R3;Indicate the three-dimensional space distance between two joint point R3 and R4;Indicate the three-dimensional space between two joint point R4 and C3
Between distance.
In above-mentioned steps six, the characteristics of human body using Kinect sensor capture current persons, by human body spy
Sign is made comparisons with the registration information of accredited personnel in database to determine the identity of current persons, flow chart as shown in Figure 2, specifically
The following steps are included:
(1) the Human Height information and the colour of skin/color development information of current persons are obtained using Kinect sensor;
(2) pass through the height for calculating and obtaining current persons according to the Human Height information of current persons, inquire database
The registration information of middle accredited personnel simultaneously traverses corresponding [h-3 Δ h, the h+3 Δ h] range of each accredited personnel, judges whether to deposit
In the accredited personnel to match with current persons' height, if there is and there is uniqueness to be then directly identified as current persons pair
The accredited personnel answered terminates to identify and exports result;Following third step is then carried out if there is but without uniqueness;If no
There are matched accredited personnel then to carry out following 4th step;Wherein h indicates that the height average value of accredited personnel, Δ h indicate standard
Difference;
(3) according to the colour of skin of current persons/color development information, the colour of skin/color development mixed Gauss model of current persons is obtained,
According to there is the accredited personnel to match with current persons' height but not unique condition on the basis of step (2), determining and waiting
Select registrant's range and judge whether there is with the colour of skin of current persons/hair color model unique match accredited personnel, if deposited
Current persons are then identified as the accredited personnel in the accredited personnel of unique match, terminates to identify and exports result;If do not deposited
4th step below the accredited personnel of unique match is then progressive;
(4) phonetic order is issued, it is desirable that front face towards Kinect sensor, and is obtained and work as forefathers by current persons
The human face image information of member;
(5) according to the human face image information of current persons, the accredited personnel's information inquired in database simultaneously judges whether to deposit
In matched accredited personnel, and if so, current persons are identified as corresponding accredited personnel, terminate to identify and export as a result,
Current persons are then identified as strange personnel or current persons are required to re-register by accredited personnel if there is no match;
It should be noted that obtaining Human Height information, the colour of skin/color development information of current persons using Kinect sensor
With the implementation of human face image information, the body height information of accredited personnel, skin are obtained using Kinect sensor with when registration
Color/color development information is identical with the implementation of human face image information.
Above embodiments are only the descriptions carried out to the preferred embodiment of the present invention, and model not is claimed to the present invention
The restriction for enclosing progress, under the premise of not departing from design principle of the present invention and spirit, this field engineers and technicians are according to this hair
The various forms of deformations that bright technical solution is made, should all fall into protection scope determined by claims of the present invention.
Claims (3)
1. a kind of identity integrated recognition method based on Kinect sensor, which is characterized in that including registration process and identified
Journey, and specifically includes the following steps:
One, it allows accredited personnel's multi-angle rotation face before Kinect sensor, and does different limb actions in different location,
To obtain the multiple groups characteristics of human body of the accredited personnel, every group of characteristics of human body include human face image information, the colour of skin/color development information and
Human Height information;The colour of skin/color development the information for obtaining accredited personnel is realized by mode in detail below:
(1) the human body head image of accredited personnel is converted into YCbCr colour gamut from RGB color domain, and in human body head image
Each pixel judge whether its CbCr chrominance component belongs to basic skin distribution U (Cb, Cr), if belonging to be labeled as 1,
0 is labeled as if being not belonging to;
(2) it according to the judgement of step (1) and label as a result, mark the pixel for being to gather as one for all, and calculates
The corresponding covariance matrix of mean value and CbCr of CbCr chrominance component, as colour of skin list Gauss model, wherein CbCr chrominance component
Mean value is usedIt indicates, covariance matrix σ1 2It indicates, colour of skin list Gauss model N1(μ1, σ1 2) indicate;
(3) it according to the judgement of step (1) and label as a result, mark the pixel for being to gather as one for all, and calculates
The corresponding covariance matrix of mean value and CbCr of CbCr chrominance component, as color development list Gauss model, wherein CbCr chrominance component
Mean value is usedIt indicates, covariance matrix σ2 2It indicates, color development list Gauss model N2(μ2, σ2 2) indicate;
The Human Height information for obtaining accredited personnel is realized by mode in detail below:
(1) artis in human skeleton is divided into five groups, the 1st group is (C1, C2, C3, C4, C5), the 2nd group for (L1, L2,
L3, L4), the 3rd group is (R1, R2, R3, R4), and the 4th group is (E1, E2, E3, E4), and the 5th group is (F1, F2, F3, F4);
(2) least square method fitting three-dimensional space straight line is respectively adopted to each group joint point set, and it is quasi- to calculate respective straight line
Error is closed, is denoted as Δ 1, Δ 2, Δ 3, Δ 4, Δ 5 respectively;
(3) when all error deltas 1, Δ 2, Δ 3, Δ 4, Δ 5 are respectively less than given threshold T1, then it is assumed that human body is in each joint
Straight configuration, and calculate the Human Height indicated with H as follows;
α=(Δ4+Δ5)/(Δ1+Δ2+Δ3+Δ4+Δ5) (6)
H=α (H1+max(H2, H3))+2(1-α)max(A1+A2) (7)
In above-mentioned formula (1) into (5),Indicate the three-dimensional space distance between two joint point C1 and C2;Indicate two
Three-dimensional space distance between artis C2 and C3;Indicate the three-dimensional space distance between two joint point C3 and C4;
Indicate the three-dimensional space distance between two joint point C4 and C5;Indicate two joint point E1 and E2 between three-dimensional space away from
From;Indicate the three-dimensional space distance between two joint point E2 and E3;Indicate the three-dimensional between two joint point E3 and E4
Space length;Indicate the three-dimensional space distance between two joint point F1 and F2;It indicates between two joint point F2 and F3
Three-dimensional space distance;Indicate the three-dimensional space distance between two joint point F3 and F4;Indicate two joint point L1 and
Three-dimensional space distance between L2;Indicate the three-dimensional space distance between two joint point L2 and L3;Indicate two joint
Three-dimensional space distance between point L3 and L4;Indicate the three-dimensional space distance between two joint point L4 and C3;Table
Show the three-dimensional space distance between two joint point R1 and R2;Indicate the three-dimensional space distance between two joint point R2 and R3;Indicate the three-dimensional space distance between two joint point R3 and R4;Indicate the three-dimensional space between two joint point R4 and C3
Between distance;
The human face image information for obtaining accredited personnel is realized by mode in detail below:
(1) using depth image and color image of the Kinect sensor acquisition comprising accredited personnel, and according in depth image
Depth data reduction accredited personnel human skeleton artis information, wherein
Torso portion includes the crown, lower jaw, chest, abdomen, hip, is successively indicated with C1, C2, C3, C4, C5;
Left-hand part includes left hand finger tip, left finesse, left elbow joint, left shoulder joint, is successively indicated with L1, L2, L3, L4;
Right hand portion includes right hand finger tip, right finesse, right elbow joint, right shoulder joint, is successively indicated with R1, R2, R3, R4;
Left leg section includes left foot point, left foot wrist, left knee joint, left hip joint, is successively indicated with E1, E2, E3, E4;
Right leg section includes right crus of diaphragm point, right crus of diaphragm wrist, right knee joint, right hip joint, is successively indicated with F1, F2, F3, F4;
(2) it using the line of two artis of C1 in the human skeleton of accredited personnel and C2 as axis, is mentioned using human body segmentation's method
The human body head region in color image is taken, as human body head image;
(3) judge whether human body head image includes face using face recognition algorithms, grab face figure if including face
Picture, as the human face image information of accredited personnel, otherwise it is assumed that not including face;
Two, the multiple groups human face image information based on the accredited personnel is extracted Haar-Like feature and is individually instructed by SVM algorithm
Practice recognition of face classifier, to obtain the recognition of face classifier result of the accredited personnel;
Three, the multiple groups colour of skin/color development information based on the accredited personnel, it is mixed by the accumulative colour of skin/color development for obtaining the accredited personnel
Close Gauss model;
Four, the multiple groups Human Height information based on the accredited personnel, by calculate obtain the accredited personnel height average value and
Standard deviation;
Five, by Step 2: the obtained result deposit database of step 3 and step 4 to complete the information registering of the accredited personnel,
And the information registering of all accredited personnel is completed according to the logon mode of the accredited personnel;
Six, after the completion of registering, using the characteristics of human body of Kinect sensor capture current persons, by the characteristics of human body of current persons
It makes comparisons with the registration information of accredited personnel in database, and determines the identity of current persons according to comparison result.
2. a kind of identity integrated recognition method based on Kinect sensor according to claim 1, it is characterised in that:
In step 3, the colour of skin/color development mixed Gauss model is N=(μ1, σ1 2, μ2, σ2 2)。
3. a kind of identity integrated recognition method based on Kinect sensor according to claim 1, it is characterised in that:
In step 6, the characteristics of human body using Kinect sensor capture current persons will infuse in the people's body characteristics and database
The registration information of volume personnel is made comparisons to determine the identity of current persons, specifically includes the following steps:
(1) the Human Height information and the colour of skin/color development information of current persons are obtained using Kinect sensor;
(2) pass through the height for calculating and obtaining current persons according to the Human Height information of current persons, inquire in database and infuse
Volume personnel registration information simultaneously traverse corresponding [h-3 Δ h, the h+3 Δ h] range of each accredited personnel, judge whether there is with
Current persons are then directly identified as corresponding by the accredited personnel that current persons' height matches if there is and with uniqueness
Accredited personnel terminates to identify and exports result;Following third step is then carried out if there is but without uniqueness;If there is no
Matched accredited personnel then carries out following 4th step;Wherein h indicates that the height average value of accredited personnel, Δ h indicate standard deviation;
(3) according to the colour of skin of current persons/color development information, the colour of skin/color development mixed Gauss model of current persons is obtained, in step
Suddenly according to there is the accredited personnel to match with current persons' height but not unique condition on the basis of (two), candidate note is determined
Volume personnel's range and judge whether there is with the colour of skin of current persons/hair color model unique match accredited personnel, if there is
Current persons are then identified as the accredited personnel by the accredited personnel of unique match, are terminated to identify and are exported result;If there is no
The accredited personnel of unique match then following 4th step of progress;
(4) phonetic order is issued, it is desirable that front face towards Kinect sensor, and is obtained current persons' by current persons
Human face image information;
(5) according to the human face image information of current persons, accredited personnel's information for inquiring in database is simultaneously judged whether there is
The accredited personnel matched, and if so, current persons are identified as corresponding accredited personnel, terminate to identify and export as a result, if
Current persons are then identified as strange personnel there is no matched accredited personnel or current persons is required to re-register;
Wherein, believed using the Human Height information, the colour of skin/color development information and facial image that Kinect sensor obtains current persons
The implementation of breath is identical as implementation when registration.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510862672.8A CN105447466B (en) | 2015-12-01 | 2015-12-01 | A kind of identity integrated recognition method based on Kinect sensor |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510862672.8A CN105447466B (en) | 2015-12-01 | 2015-12-01 | A kind of identity integrated recognition method based on Kinect sensor |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105447466A CN105447466A (en) | 2016-03-30 |
CN105447466B true CN105447466B (en) | 2019-07-23 |
Family
ID=55557626
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510862672.8A Active CN105447466B (en) | 2015-12-01 | 2015-12-01 | A kind of identity integrated recognition method based on Kinect sensor |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105447466B (en) |
Families Citing this family (24)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105773633B (en) * | 2016-04-14 | 2018-04-20 | 中南大学 | Mobile robot man-machine control system based on face location and sensitivity parameter |
JP6688990B2 (en) * | 2016-04-28 | 2020-04-28 | パナソニックIpマネジメント株式会社 | Identification device, identification method, identification program, and recording medium |
CN106599785B (en) * | 2016-11-14 | 2020-06-30 | 深圳奥比中光科技有限公司 | Method and equipment for establishing human body 3D characteristic identity information base |
CN106652291A (en) * | 2016-12-09 | 2017-05-10 | 华南理工大学 | Indoor simple monitoring and alarming system and method based on Kinect |
CN106778615B (en) * | 2016-12-16 | 2019-10-18 | 中新智擎科技有限公司 | A kind of method, apparatus and service for infrastructure robot identifying user identity |
CN106934377B (en) * | 2017-03-14 | 2020-03-17 | 新疆智辰天林信息科技有限公司 | Improved human face detection system |
TWI604332B (en) * | 2017-03-24 | 2017-11-01 | 緯創資通股份有限公司 | Method, system, and computer-readable recording medium for long-distance person identification |
CN107192342A (en) * | 2017-05-11 | 2017-09-22 | 广州帕克西软件开发有限公司 | A kind of measuring method and system of contactless build data |
CN107292252B (en) * | 2017-06-09 | 2020-09-15 | 南京华捷艾米软件科技有限公司 | Identity recognition method for autonomous learning |
CN109426785B (en) * | 2017-08-31 | 2021-09-10 | 杭州海康威视数字技术股份有限公司 | Human body target identity recognition method and device |
CN109426787A (en) * | 2017-08-31 | 2019-03-05 | 杭州海康威视数字技术股份有限公司 | A kind of human body target track determines method and device |
CN108451534B (en) * | 2018-01-26 | 2021-08-27 | 仰人杰 | Human body motion detection method based on dielectric elastomer sensor |
CN108734083B (en) * | 2018-03-21 | 2023-04-25 | 北京猎户星空科技有限公司 | Control method, device, equipment and storage medium of intelligent equipment |
CN108960078A (en) * | 2018-06-12 | 2018-12-07 | 温州大学 | A method of based on monocular vision, from action recognition identity |
CN109272347A (en) * | 2018-08-16 | 2019-01-25 | 苏宁易购集团股份有限公司 | A kind of statistical analysis technique and system of shops's volume of the flow of passengers |
CN110594856A (en) * | 2019-08-12 | 2019-12-20 | 青岛经济技术开发区海尔热水器有限公司 | Hot water circulation control method and hot water system |
CN110503022A (en) * | 2019-08-19 | 2019-11-26 | 北京积加科技有限公司 | A kind of personal identification method, apparatus and system |
CN110524559B (en) * | 2019-08-30 | 2022-06-10 | 成都未至科技有限公司 | Intelligent man-machine interaction system and method based on personnel behavior data |
CN111064925B (en) * | 2019-12-04 | 2021-05-04 | 常州工业职业技术学院 | Subway passenger ticket evasion behavior detection method and system |
CN111079644B (en) * | 2019-12-13 | 2023-06-06 | 四川新网银行股份有限公司 | Method for assisting photographing based on distance and joint point identification external force and storage medium |
CN111199198B (en) * | 2019-12-27 | 2023-08-04 | 深圳市优必选科技股份有限公司 | Image target positioning method, image target positioning device and mobile robot |
CN111292087A (en) * | 2020-01-20 | 2020-06-16 | 北京沃东天骏信息技术有限公司 | Identity verification method and device, computer readable medium and electronic equipment |
CN114582003B (en) * | 2022-04-24 | 2022-07-29 | 慕思健康睡眠股份有限公司 | Sleep health management system based on cloud computing service |
CN115637901A (en) * | 2022-10-09 | 2023-01-24 | 东风汽车集团股份有限公司 | Child lock control system, method and equipment based on OMS |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103529944A (en) * | 2013-10-17 | 2014-01-22 | 合肥金诺数码科技股份有限公司 | Human body movement identification method based on Kinect |
CN103606093A (en) * | 2013-10-28 | 2014-02-26 | 燕山大学 | Intelligent chain VIP customer service system based on human characteristics |
CN104167016A (en) * | 2014-06-16 | 2014-11-26 | 西安工业大学 | Three-dimensional motion reconstruction method based on RGB color and depth image |
CN104766230A (en) * | 2015-04-21 | 2015-07-08 | 东华大学 | Advertising effect evaluation method based on human skeletal tracking |
-
2015
- 2015-12-01 CN CN201510862672.8A patent/CN105447466B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103529944A (en) * | 2013-10-17 | 2014-01-22 | 合肥金诺数码科技股份有限公司 | Human body movement identification method based on Kinect |
CN103606093A (en) * | 2013-10-28 | 2014-02-26 | 燕山大学 | Intelligent chain VIP customer service system based on human characteristics |
CN104167016A (en) * | 2014-06-16 | 2014-11-26 | 西安工业大学 | Three-dimensional motion reconstruction method based on RGB color and depth image |
CN104766230A (en) * | 2015-04-21 | 2015-07-08 | 东华大学 | Advertising effect evaluation method based on human skeletal tracking |
Also Published As
Publication number | Publication date |
---|---|
CN105447466A (en) | 2016-03-30 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105447466B (en) | A kind of identity integrated recognition method based on Kinect sensor | |
CN106529468B (en) | A kind of finger vein identification method and system based on convolutional neural networks | |
CN105574518B (en) | Method and device for detecting living human face | |
US6920236B2 (en) | Dual band biometric identification system | |
CN103914699B (en) | A kind of method of the image enhaucament of the automatic lip gloss based on color space | |
CN101561710B (en) | Man-machine interaction method based on estimation of human face posture | |
CN106778785B (en) | Construct the method for image Feature Selection Model and the method, apparatus of image recognition | |
Roomi et al. | Race classification based on facial features | |
CN109543640A (en) | A kind of biopsy method based on image conversion | |
Tayal et al. | Automatic face detection using color based segmentation | |
CN109558825A (en) | A kind of pupil center's localization method based on digital video image processing | |
Shangeetha et al. | Computer vision based approach for Indian Sign Language character recognition | |
CN109063671A (en) | Method and device for intelligent cosmetic | |
CN103984922B (en) | Face identification method based on sparse representation and shape restriction | |
CN107392151A (en) | Face image various dimensions emotion judgement system and method based on neutral net | |
CN109002799A (en) | Face identification method | |
CN109766782A (en) | Real-time body action identification method based on SVM | |
CN104898971B (en) | A kind of mouse pointer control method and system based on Visual Trace Technology | |
Seal et al. | Minutiae based thermal face recognition using blood perfusion data | |
Akhloufi et al. | Thermal faceprint: A new thermal face signature extraction for infrared face recognition | |
CN114863499A (en) | Finger vein and palm vein identification method based on federal learning | |
Sokhib et al. | A combined method of skin-and depth-based hand gesture recognition. | |
Akhloufi et al. | Infrared face recognition using distance transforms | |
CN108288040A (en) | Multi-parameter face identification system based on face contour | |
CN109711306A (en) | A kind of method and apparatus obtaining facial characteristics based on depth convolutional neural networks |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |