CN107590452A - A kind of personal identification method and device based on gait and face fusion - Google Patents

A kind of personal identification method and device based on gait and face fusion Download PDF

Info

Publication number
CN107590452A
CN107590452A CN201710786656.4A CN201710786656A CN107590452A CN 107590452 A CN107590452 A CN 107590452A CN 201710786656 A CN201710786656 A CN 201710786656A CN 107590452 A CN107590452 A CN 107590452A
Authority
CN
China
Prior art keywords
human
quality evaluation
image
gait
facial 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.)
Pending
Application number
CN201710786656.4A
Other languages
Chinese (zh)
Inventor
姚琪
刘靖峰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Wuhan Vision Information Technology Co Ltd
Original Assignee
Wuhan Vision Information Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Wuhan Vision Information Technology Co Ltd filed Critical Wuhan Vision Information Technology Co Ltd
Priority to CN201710786656.4A priority Critical patent/CN107590452A/en
Publication of CN107590452A publication Critical patent/CN107590452A/en
Pending legal-status Critical Current

Links

Abstract

The present invention relates to a kind of gait based on deep learning and face fusion personal identification method and device, wherein device includes video acquisition and pretreatment module, Method of Gait Feature Extraction module, face characteristic extraction module and identification module, method includes collection video flowing, pedestrian detection and tracking, and Face datection are carried out to video flowing;Method of Gait Feature Extraction is carried out to the human body image, and calculates the quality evaluation fraction of the gait feature;Face characteristic extraction is carried out to facial image, and calculates the quality evaluation fraction of face characteristic, using quality evaluation fraction highest facial image as facial image to be identified;Gait feature and face characteristic are weighted according to respective quality score, input SVM classifier carries out identification.By the quality evaluation fraction for calculating gait feature and face characteristic, and be weighted according to quality score, combine the advantage of recognition of face and Gait Recognition, the complementation of two kinds of identification technologies, the robustness of identifying system is added, improves the accuracy of piece identity's identification.

Description

A kind of personal identification method and device based on gait and face fusion
Technical field
The present invention relates to biometrics identification technology field, and in particular to a kind of to be known based on the identity of gait and face fusion Other method and device.
Background technology
Biometrics identification technology is a kind of technology that identification is carried out using human body biological characteristics.These biological characteristics Including:Fingerprint, palmmprint, iris, face etc., extensively using various peaces such as security protection, bank, customs's frontier inspection, estate managements In full field, the accuracy rate for improving identification greatly reduces man power and material again, is increasingly favored by various circles of society.
Recognition of face is a kind of contactless biometric identification, it is not necessary to which the behavior of people coordinates, especially suitable for long distance From identification, can be used for intelligent video monitoring system.Current face recognition technology, when target person side is to camera When, camera can only collect the side face of target person, and recognition accuracy can significantly decline.On the other hand, according to people Posture on foot identifies the technology of target person identity (being also Gait Recognition), has non-contact remote and is not easy to pretend The features such as, increasing researcher's concern has been obtained in recent years.But when target person face camera, due to The posture of target person can not be obtained well, and recognition effect is had a greatly reduced quality.
The content of the invention
The present invention is directed to technical problem of the prior art, there is provided a kind of identification side based on gait and face fusion Method and device, by combining Gait Recognition and recognition of face, complementation is carried out to two kinds of identification technologies, to improve piece identity's identification Accuracy.
The technical scheme that the present invention solves above-mentioned technical problem is as follows:
On the one hand, the present invention provides a kind of personal identification method based on gait and face fusion, comprises the following steps:
S1, video flowing is gathered, pedestrian detection and tracking are carried out to video flowing, obtain human body frame image sequence, while to people Body two field picture carries out Face datection, obtains human face image sequence corresponding to human body frame image sequence;
S2, Method of Gait Feature Extraction is carried out to the human body image in the human body frame image sequence, and obtain quality evaluation ginseng Number, the quality evaluation fraction of the gait feature is calculated according to quality evaluation parameter;
S3, face characteristic extraction is carried out to each image in human face image sequence, and obtain the quality evaluation of facial image Parameter, the quality evaluation fraction of face characteristic is calculated according to quality evaluation parameter, by quality evaluation fraction highest facial image As facial image to be identified;
S4, gait feature and face characteristic are weighted according to respective quality score, input SVM classifier carries out body Part identification.
The beneficial effects of the invention are as follows:By calculating the quality evaluation fraction of gait feature and face characteristic, and gait Feature and face characteristic are weighted according to respective quality score, combine the advantage of recognition of face and Gait Recognition, two kinds The complementation of identification technology, the robustness of identifying system is added, improve the accuracy of piece identity's identification.
On the basis of above-mentioned technical proposal, the present invention can also do following improvement.
Further, pedestrian detection is carried out to video flowing described in step S1, using Faster-RCNN deep learning side Method is realized, while uses correlation filtering, to detecting that pedestrian is tracked;The Face datection uses SSD single detectors Capture face.
It is using the above-mentioned further beneficial effect of scheme:Faster-RCNN deep learning method is than common use Edge feature, shape facility, statistical nature etc. are more accurate, there has also been very big lifting in processing ground speed, it is possible to achieve real When the demand that handles.Correlation filtering is used simultaneously, to detecting that pedestrian is tracked, with the real-time of lifting processing.SSD (Single Shot MultiBox Detector) is to realize Target detection and identification using single deep neural network model Method.Combine Faster R-CNN anchor and YOLO single Neurals detection thinking, it is possible to achieve high-accuracy is examined in real time Survey.
Preferably, the step S2 includes following sub-step:
S21, size normalization processing is carried out to the human body image in human body frame image sequence;
S22, the human body frame image sequence input gestures after processing are identified into network, extract the posture of each human body image Feature, posture characteristic sequence corresponding to human body frame image sequence is generated, while extract the quality evaluation parameter of human body two field picture, and The quality evaluation fraction of human body two field picture is calculated according to quality evaluation parameter;
S23, the posture characteristic sequence is inputted into Gait Recognition network, obtain for describe the feature of body gait to Amount.
Preferably, the step S3, including following sub-step:
S31, Face datection is carried out using SSD singles detector to each facial image in human face image sequence, carried The 3D Landmark of every face are taken, while obtain the quality evaluation parameter of facial image, and are commented according to the quality of facial image Valency parameter calculates the quality evaluation fraction of every facial image;
S32, choose human face image sequence in quality evaluation fraction highest facial image as facial image to be identified, and Three-dimensional modeling is carried out according to the 3D Landmark of the facial image to be identified, projects front face;
S33, the characteristic vector of facial image to be identified is obtained by face characteristic neutral net.
Further, the quality evaluation parameter includes size, angle and the definition of human body image or facial image.
On the other hand, the present invention provides a kind of identity recognition device based on gait and face fusion, including:
Video acquisition and pretreatment module, for gathering video flowing, pedestrian detection and tracking are carried out to video flowing, obtain people Body frame image sequence, while Face datection is carried out to human body two field picture, obtain facial image sequence corresponding to human body frame image sequence Row;
Method of Gait Feature Extraction module, Method of Gait Feature Extraction is carried out to the human body image in the human body frame image sequence, and Quality evaluation parameter is obtained, the quality evaluation fraction of the gait feature is calculated according to quality evaluation parameter;
Face characteristic extraction module, face characteristic extraction is carried out to each image in human face image sequence, and obtain face The quality evaluation parameter of image, the quality evaluation fraction of face characteristic is calculated according to quality evaluation parameter, by quality evaluation fraction Highest facial image is as facial image to be identified;
Identification module, gait feature and face characteristic are weighted according to respective quality score, input SVM classifier Carry out identification.
The beneficial effects of the invention are as follows:By calculating the quality evaluation fraction of gait feature and face characteristic, and gait Feature and face characteristic are weighted according to respective quality score, combine the advantage of recognition of face and Gait Recognition, two kinds The complementation of identification technology, the robustness of identifying system is added, improve the accuracy of piece identity's identification.
Further, the video acquisition and pretreatment module use Faster-RCNN deep learning method to video flowing Pedestrian detection is carried out, while uses correlation filtering, to detecting that pedestrian is tracked;Face is carried out using SSD singles detector Detection, capture face.
Beneficial effect using above-mentioned further scheme is that Faster-RCNN deep learning method is than common use Edge feature, shape facility, statistical nature etc. are more accurate, there has also been very big lifting in processing ground speed, it is possible to achieve real When the demand that handles.Correlation filtering is used simultaneously, to detecting that pedestrian is tracked, with the real-time of lifting processing.SSD (Single Shot MultiBox Detector) is to realize Target detection and identification using single deep neural network model Method.Combine Faster R-CNN anchor and YOLO single Neurals detection thinking, it is possible to achieve high-accuracy is examined in real time Survey.
Preferably, the Method of Gait Feature Extraction module is specifically used for:
Size normalization processing is carried out to the human body image in human body frame image sequence;
Human body frame image sequence input gestures after processing are identified into network, the posture for extracting each human body image is special Sign, posture characteristic sequence corresponding to human body frame image sequence is generated, while extract the quality evaluation parameter of human body two field picture, and root The quality evaluation fraction of human body two field picture is calculated according to quality evaluation parameter;
The posture characteristic sequence is inputted into Gait Recognition network, obtains the characteristic vector for describing body gait.
Preferably, the face characteristic extraction module is specifically used for:
Face datection is carried out using SSD singles detector to each facial image in human face image sequence, extraction is every The 3D Landmark of face are opened, while obtain the quality evaluation parameter of facial image, and are joined according to the quality evaluation of facial image Number calculates the quality evaluation fraction of every facial image;
Choose human face image sequence in quality evaluation fraction highest facial image as facial image to be identified, and according to The 3D Landmark of the facial image to be identified carry out three-dimensional modeling, project front face;
The characteristic vector of facial image to be identified is obtained by face characteristic neutral net.
Preferably, the quality evaluation parameter includes size, angle and the definition of human body image or facial image.
Brief description of the drawings
Fig. 1 is a kind of personal identification method flow chart based on gait and face fusion provided in an embodiment of the present invention;
Fig. 2 is gait feature abstracting method flow chart provided in an embodiment of the present invention;
Fig. 3 is face feature extraction method flow chart provided in an embodiment of the present invention;
Fig. 4 is weighted calculation provided in an embodiment of the present invention and personal identification method flow chart;
Fig. 5 is a kind of identity recognition device structure chart based on gait and face fusion provided in an embodiment of the present invention;
Embodiment
The principle and feature of the present invention are described below in conjunction with example, the given examples are served only to explain the present invention, and It is non-to be used to limit the scope of the present invention.
Fig. 1 is a kind of personal identification method flow chart based on gait and face fusion provided in an embodiment of the present invention.Such as Shown in Fig. 1, comprise the following steps:
S1, video flowing is gathered, pedestrian detection and tracking are carried out to video flowing, obtain human body frame image sequence, while to people Body two field picture carries out Face datection, obtains human face image sequence corresponding to human body frame image sequence;
S2, Method of Gait Feature Extraction is carried out to the human body image in the human body frame image sequence, and obtain quality evaluation ginseng Number, the quality evaluation fraction of the gait feature is calculated according to quality evaluation parameter;Referring specifically to Fig. 2;
S3, face characteristic extraction is carried out to each image in human face image sequence, and obtain the quality evaluation of facial image Parameter, the quality evaluation fraction of face characteristic is calculated according to quality evaluation parameter, by quality evaluation fraction highest facial image As facial image to be identified;Referring specifically to Fig. 3;
S4, gait feature and face characteristic are weighted according to respective quality score, input SVM classifier carries out body Part identification, referring specifically to Fig. 4.
It is the angle of the quality evaluation parameter including image, size, definition, contrast, bright in above-described embodiment The factor parameter of picture quality is influenceed as defined in the ISO/IEC19794-5 such as degree, definition standards, by calculate gait feature and The quality evaluation fraction of face characteristic, and gait feature and face characteristic are weighted according to respective quality score, it is comprehensive The advantage of recognition of face and Gait Recognition, the complementations of two kinds of identification technologies, the robustness of identifying system is added, improves personage The accuracy of identification.According to ISO/IEC19794-5 standards, the factor that influences picture quality is fuzzy, non-frontal image, Side illumination, by a variety of evaluation indexes such as the contrast of image, lightness, definition, the size of face, position and angles Carry out overall merit facial image, draw respective evaluation coefficient (0~1.0), then by each coefficient proportion (1~ 100) it is weighted, finally draws the evaluation score of every width figure, so as to selects the optimal facial image of quality.
On the basis of above-described embodiment, pedestrian detection is carried out to video flowing described in step S1, using Faster- RCNN deep learning method is realized, while uses correlation filtering, to detecting that pedestrian is tracked;The Face datection is adopted Face is captured with SSD singles detector.
In above-described embodiment, Faster-RCNN deep learning method than it is common using edge feature, shape facility, Statistical nature etc. is more accurate, there has also been very big lifting in processing ground speed, it is possible to achieve the demand handled in real time.Adopt simultaneously With correlation filtering, to detecting that pedestrian is tracked, with the real-time of lifting processing.SSD(Single Shot MultiBox Detector) it is method that Target detection and identification is realized using single deep neural network model.Combine Faster R- CNN anchor and YOLO single Neurals detection thinking, it is possible to achieve high-accuracy detects in real time.
Preferably, the step S2 includes following sub-step, as shown in Fig. 2:
S21, size normalization processing is carried out to the human body image in human body frame image sequence;
S22, the human body frame image sequence input gestures after processing are identified into network, extract the posture of each human body image Feature, posture characteristic sequence corresponding to human body frame image sequence is generated, while extract the quality evaluation parameter of human body two field picture, and The quality evaluation fraction of human body two field picture is calculated according to quality evaluation parameter;
S23, the posture characteristic sequence is inputted into Gait Recognition network, obtain for describe the feature of body gait to Amount.
Preferably, the step S3, including following sub-step, as shown in Figure 3:
S31, Face datection is carried out using SSD singles detector to each facial image in human face image sequence, carried The 3D Landmark of every face are taken, while obtain the quality evaluation parameter of facial image, and are commented according to the quality of facial image Valency parameter calculates the quality evaluation fraction of every facial image;
S32, choose human face image sequence in quality evaluation fraction highest facial image as facial image to be identified, and Three-dimensional modeling is carried out according to the 3D Landmark of the facial image to be identified, projects front face;
S33, the characteristic vector of facial image to be identified is obtained by face characteristic neutral net.
Preferably, the quality evaluation parameter includes size, angle and the definition of human body image or facial image.
On the other hand, the present invention provides a kind of identity recognition device based on gait and face fusion, as shown in figure 5, bag Include:
Video acquisition and pretreatment module, for gathering video flowing, pedestrian detection and tracking are carried out to video flowing, obtain people Body frame image sequence, while Face datection is carried out to human body two field picture, obtain facial image sequence corresponding to human body frame image sequence Row;
Method of Gait Feature Extraction module, Method of Gait Feature Extraction is carried out to the human body image in the human body frame image sequence, and Quality evaluation parameter is obtained, the quality evaluation fraction of the gait feature is calculated according to quality evaluation parameter;
Face characteristic extraction module, face characteristic extraction is carried out to each image in human face image sequence, and obtain face The quality evaluation parameter of image, the quality evaluation fraction of face characteristic is calculated according to quality evaluation parameter, by quality evaluation fraction Highest facial image is as facial image to be identified;
Identification module, gait feature and face characteristic are weighted according to respective quality score, input SVM classifier Carry out identification.
In above-described embodiment, by calculating the quality evaluation fraction of gait feature and face characteristic, and gait feature and Face characteristic is weighted according to respective quality score, combines the advantage of recognition of face and Gait Recognition, two kinds of identification skills The complementation of art, the robustness of identifying system is added, improve the accuracy of piece identity's identification.Marked according to ISO/IEC19794-5 Standard, the factor for influenceing quality of human face image is fuzzy, non-frontal image, side illumination, by the contrast to image, is become clear A variety of evaluation indexes such as degree, definition, the size of face, position and angle carry out overall merit facial image, draw respective comment Valency coefficient (0~1.0), then it is weighted by each coefficient proportion (1~100), finally draws every width figure Score is evaluated, so as to select the optimal facial image of quality.
On the basis of above-described embodiment, the video acquisition and pretreatment module use Faster-RCNN depth Learning method carries out pedestrian detection to video flowing, while uses correlation filtering, to detecting that pedestrian is tracked;Using SSD singles Detector carries out Face datection, captures face.
In above-described embodiment, Faster-RCNN deep learning method than it is common using edge feature, shape facility, Statistical nature etc. is more accurate, there has also been very big lifting in processing ground speed, it is possible to achieve the demand handled in real time.Adopt simultaneously With correlation filtering, to detecting that pedestrian is tracked, with the real-time of lifting processing.SSD(Single Shot MultiBox Detector) it is method that Target detection and identification is realized using single deep neural network model.Combine Faster R- CNN anchor and YOLO single Neurals detection thinking, it is possible to achieve high-accuracy detects in real time.
Preferably, the Method of Gait Feature Extraction module is specifically used for:
Size normalization processing is carried out to the human body image in human body frame image sequence;
Human body frame image sequence input gestures after processing are identified into network, the posture for extracting each human body image is special Sign, posture characteristic sequence corresponding to human body frame image sequence is generated, while extract the quality evaluation parameter of human body two field picture, and root The quality evaluation fraction of human body two field picture is calculated according to quality evaluation parameter;
The posture characteristic sequence is inputted into Gait Recognition network, obtains the characteristic vector for describing body gait.
Preferably, the face characteristic extraction module is specifically used for:
Face datection is carried out using SSD singles detector to each facial image in human face image sequence, extraction is every The 3D Landmark of face are opened, while obtain the quality evaluation parameter of facial image, and are joined according to the quality evaluation of facial image Number calculates the quality evaluation fraction of every facial image;
Choose human face image sequence in quality evaluation fraction highest facial image as facial image to be identified, and according to The 3D Landmark of the facial image to be identified carry out three-dimensional modeling, project front face;
The characteristic vector of facial image to be identified is obtained by face characteristic neutral net.
Preferably, the quality evaluation parameter includes size, angle and the definition of human body image or facial image.
The gesture recognition model that we are obtained with 2 kinds of differences based on different pieces of information collection (PFM and H3.6M) training respectively, with And same Gait Recognition model, do 2 kinds of tests.The first test, is determined with the sequence for containing 10 continuous postures A certain personal gait, 305 people (wherein 100 people are used to train, and 150 people are used to verify, 155 are used to test) are shared, everyone Clothing is consistent, but has three kinds of walking manners, including:Normal walking, knapsack and wears rubber overshoes.
The Gait Recognition accuracy rate of the test of table 11
Normal walking Knapsack Wear rubber overshoes It is average
H3.6M 99.4% 92.6% 98.7% 96.9%
PFM 99.7% 99.0% 99.0% 99.2%
Second of test, determines a certain personal gait with the sequence for containing 20 continuous postures, shares 32 people (10 Individual is used for training pattern, and 6 people are used to verify, 16 people are used to test), everyone wears Various Seasonal, equally including three Kind walking manner.
The Gait Recognition accuracy rate of the test of table 22
Normal walking Knapsack Wear rubber overshoes It is average
H3.6M 74.2% 89.1% 82.8% 80.1%
PFM 78.1% 56.3% 46.9% 60.4%
In terms of the test result of Tables 1 and 2, the gait for capturing pedestrian by depth learning technology has certain robust Property, obtained attitude mode is especially trained by H3.6M data sets, is more suitable for that the pedestrian of different garment is identified.
Finally recognition of face and Gait Recognition are combined mode by us, go to test 32 people above again, finally Test result is as shown in table 3, and the mode combined with both is significantly lifted to identify that pedestrian has in accuracy rate.
The recognition of face of table 3 and the accuracy rate of Gait Recognition
Normal walking Knapsack Wear rubber overshoes It is average
H3.6M 97.2% 98.5% 99.0% 98.2%
PFM 95.1% 89.3% 85.9% 90.1%
The foregoing is only presently preferred embodiments of the present invention, be not intended to limit the invention, it is all the present invention spirit and Within principle, any modification, equivalent substitution and improvements made etc., it should be included in the scope of the protection.

Claims (10)

1. a kind of personal identification method based on gait and face fusion, it is characterised in that comprise the following steps:
S1, video flowing is gathered, pedestrian detection and tracking are carried out to video flowing, obtain human body frame image sequence, while to human body frame Image carries out Face datection, obtains human face image sequence corresponding to human body frame image sequence;
S2, Method of Gait Feature Extraction is carried out to the human body image in the human body frame image sequence, and obtain quality evaluation parameter, root The quality evaluation fraction of the gait feature is calculated according to quality evaluation parameter;
S3, face characteristic extraction is carried out to each image in human face image sequence, and obtains the quality evaluation parameter of facial image, The quality evaluation fraction of face characteristic is calculated according to quality evaluation parameter, using quality evaluation fraction highest facial image as treating Identify facial image;
S4, gait feature and face characteristic are weighted according to respective quality score, input SVM classifier carries out identity knowledge Not.
A kind of 2. personal identification method based on gait and face fusion according to claim 1, it is characterised in that step Pedestrian detection is carried out to video flowing described in S1, realized using Faster-RCNN deep learning method, while use phase Filtering is closed, to detecting that pedestrian is tracked;The Face datection is using SSD singles detector crawl face.
3. a kind of personal identification method based on gait and face fusion according to claim 2, it is characterised in that described Step S2 includes following sub-step:
S21, size normalization processing is carried out to the human body image in human body frame image sequence;
S22, the human body frame image sequence input gestures after processing are identified into network, the posture for extracting each human body image is special Sign, posture characteristic sequence corresponding to human body frame image sequence is generated, while extract the quality evaluation parameter of human body two field picture, and root The quality evaluation fraction of human body two field picture is calculated according to quality evaluation parameter;
S23, the posture characteristic sequence is inputted into Gait Recognition network, obtains the characteristic vector for describing body gait.
4. a kind of personal identification method based on gait and face fusion according to claim 3, it is characterised in that described Step S3, including following sub-step:
S31, Face datection is carried out using SSD singles detector to each facial image in human face image sequence, extraction is every The 3D Landmark of face are opened, while obtain the quality evaluation parameter of facial image, and are joined according to the quality evaluation of facial image Number calculates the quality evaluation fraction of every facial image;
S32, choose human face image sequence in quality evaluation fraction highest facial image as facial image to be identified, and according to The 3D Landmark of the facial image to be identified carry out three-dimensional modeling, project front face;
S33, the characteristic vector of facial image to be identified is obtained by face characteristic neutral net.
5. a kind of personal identification method based on gait and face fusion according to claim 1, it is characterised in that described Quality evaluation parameter includes size, angle and the definition of human body image or facial image.
A kind of 6. identity recognition device based on gait and face fusion, it is characterised in that including:
Video acquisition and pretreatment module, for gathering video flowing, pedestrian detection and tracking are carried out to video flowing, obtain human body frame Image sequence, while Face datection is carried out to human body two field picture, obtain human face image sequence corresponding to human body frame image sequence;
Method of Gait Feature Extraction module, Method of Gait Feature Extraction is carried out to the human body image in the human body frame image sequence, and obtained Quality evaluation parameter, the quality evaluation fraction of the gait feature is calculated according to quality evaluation parameter;
Face characteristic extraction module, face characteristic extraction is carried out to each image in human face image sequence, and obtain facial image Quality evaluation parameter, according to quality evaluation parameter calculate face characteristic quality evaluation fraction, by quality evaluation fraction highest Facial image as facial image to be identified;
Identification module, gait feature and face characteristic are weighted according to respective quality score, input SVM classifier is carried out Identification.
7. a kind of identity recognition device based on gait and face fusion according to claim 6, it is characterised in that described Video acquisition and pretreatment module carry out pedestrian detection using Faster-RCNN deep learning method to video flowing, adopt simultaneously With correlation filtering, to detecting that pedestrian is tracked;Face datection is carried out using SSD singles detector, captures face.
8. a kind of identity recognition device based on gait and face fusion according to claim 7, it is characterised in that described Method of Gait Feature Extraction module is specifically used for:
Size normalization processing is carried out to the human body image in human body frame image sequence;
Human body frame image sequence input gestures after processing are identified into network, extract the posture feature of each human body image, it is raw Posture characteristic sequence corresponding to adult body frame image sequence, while the quality evaluation parameter of human body two field picture is extracted, and according to matter Measure the quality evaluation fraction that evaluating calculates human body two field picture;
The posture characteristic sequence is inputted into Gait Recognition network, obtains the characteristic vector for describing body gait.
9. a kind of identity recognition device based on gait and face fusion according to claim 8, it is characterised in that described Face characteristic extraction module is specifically used for:
Face datection is carried out using SSD singles detector to each facial image in human face image sequence, extracts every people The 3D Landmark of face, while the quality evaluation parameter of facial image is obtained, and according to the quality evaluation parameter meter of facial image Calculate the quality evaluation fraction of every facial image;
Choose in human face image sequence that quality evaluation fraction highest facial image is as facial image to be identified, and according to described The 3D Landmark of facial image to be identified carry out three-dimensional modeling, project front face;
The characteristic vector of facial image to be identified is obtained by face characteristic neutral net.
A kind of 10. other device of identification based on gait and face fusion according to claim 6, it is characterised in that The quality evaluation parameter includes size, angle and the definition of human body image or facial image.
CN201710786656.4A 2017-09-04 2017-09-04 A kind of personal identification method and device based on gait and face fusion Pending CN107590452A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710786656.4A CN107590452A (en) 2017-09-04 2017-09-04 A kind of personal identification method and device based on gait and face fusion

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710786656.4A CN107590452A (en) 2017-09-04 2017-09-04 A kind of personal identification method and device based on gait and face fusion

Publications (1)

Publication Number Publication Date
CN107590452A true CN107590452A (en) 2018-01-16

Family

ID=61050757

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710786656.4A Pending CN107590452A (en) 2017-09-04 2017-09-04 A kind of personal identification method and device based on gait and face fusion

Country Status (1)

Country Link
CN (1) CN107590452A (en)

Cited By (28)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108108693A (en) * 2017-12-20 2018-06-01 深圳市安博臣实业有限公司 Intelligent identification monitoring device and recognition methods based on 3D high definition VR panoramas
CN108460364A (en) * 2018-03-27 2018-08-28 百度在线网络技术(北京)有限公司 Method and apparatus for generating information
CN108629299A (en) * 2018-04-24 2018-10-09 武汉幻视智能科技有限公司 A kind of matched long-time multi-object tracking method of combination face and system
CN108734107A (en) * 2018-04-24 2018-11-02 武汉幻视智能科技有限公司 A kind of multi-object tracking method and system based on face
CN109086859A (en) * 2018-08-01 2018-12-25 上海小蚁科技有限公司 People flow rate statistical method, apparatus and computer readable storage medium
CN109145742A (en) * 2018-07-19 2019-01-04 银河水滴科技(北京)有限公司 A kind of pedestrian recognition method and system
CN109325423A (en) * 2018-08-29 2019-02-12 安徽超清科技股份有限公司 A kind of optimization SSD algorithm for pedestrian detection
CN109359544A (en) * 2018-09-19 2019-02-19 武汉烽火众智数字技术有限责任公司 A kind of portrait search method and device
CN109409297A (en) * 2018-10-30 2019-03-01 咪付(广西)网络技术有限公司 A kind of personal identification method based on binary channels convolutional neural networks
CN109446936A (en) * 2018-10-12 2019-03-08 银河水滴科技(北京)有限公司 A kind of personal identification method and device for monitoring scene
CN109754461A (en) * 2018-12-29 2019-05-14 深圳云天励飞技术有限公司 Image processing method and related product
CN109753917A (en) * 2018-12-29 2019-05-14 中国科学院重庆绿色智能技术研究院 Face quality optimization method, system, computer readable storage medium and equipment
CN109920111A (en) * 2019-03-05 2019-06-21 杭州电子科技大学 A kind of access control system of recognition of face and Gait Recognition fusion
CN110175553A (en) * 2019-05-23 2019-08-27 银河水滴科技(北京)有限公司 The method and device of feature database is established based on Gait Recognition and recognition of face
CN110175587A (en) * 2019-05-30 2019-08-27 黄岩 A kind of video frequency tracking method based on recognition of face and Algorithm for gait recognition
CN110276288A (en) * 2019-06-17 2019-09-24 北京影谱科技股份有限公司 A kind of personal identification method and device based on biological characteristic
CN110287889A (en) * 2019-06-26 2019-09-27 银河水滴科技(北京)有限公司 A kind of method and device of identification
CN110443232A (en) * 2018-04-12 2019-11-12 腾讯科技(深圳)有限公司 Method for processing video frequency and relevant apparatus, image processing method and relevant apparatus
CN110569707A (en) * 2019-06-25 2019-12-13 深圳和而泰家居在线网络科技有限公司 identity recognition method and electronic equipment
CN110750671A (en) * 2019-09-05 2020-02-04 浙江省北大信息技术高等研究院 Pedestrian retrieval method and device based on massive unstructured features
CN110796094A (en) * 2019-10-30 2020-02-14 上海商汤智能科技有限公司 Control method and device based on image recognition, electronic equipment and storage medium
CN110874582A (en) * 2019-11-18 2020-03-10 浙江大华技术股份有限公司 Telecommunication fraud determination method and device, storage medium and electronic device
CN111460884A (en) * 2020-02-09 2020-07-28 天津博宜特科技有限公司 Multi-face recognition method based on human body tracking
CN111881732A (en) * 2020-06-17 2020-11-03 北京邮电大学 SVM (support vector machine) -based face quality evaluation method
WO2020249025A1 (en) * 2019-06-12 2020-12-17 杭州海康威视数字技术股份有限公司 Identity information determining method and apparatus, and storage medium
CN112287770A (en) * 2020-10-10 2021-01-29 武汉大学 Face quality sensing method and system for identity recognition
CN112364829A (en) * 2020-11-30 2021-02-12 北京有竹居网络技术有限公司 Face recognition method, device, equipment and storage medium
CN112488071A (en) * 2020-12-21 2021-03-12 重庆紫光华山智安科技有限公司 Method, device, electronic equipment and storage medium for extracting pedestrian features

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050031195A1 (en) * 2003-08-08 2005-02-10 Microsoft Corporation System and method for modeling three dimensional objects from a single image
CN101404091A (en) * 2008-11-07 2009-04-08 重庆邮电大学 Three-dimensional human face reconstruction method and system based on two-step shape modeling
AU2011101355A4 (en) * 2011-10-20 2011-12-08 Girija Chetty Biometric person identity verification base on face and gait fusion

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050031195A1 (en) * 2003-08-08 2005-02-10 Microsoft Corporation System and method for modeling three dimensional objects from a single image
CN101404091A (en) * 2008-11-07 2009-04-08 重庆邮电大学 Three-dimensional human face reconstruction method and system based on two-step shape modeling
AU2011101355A4 (en) * 2011-10-20 2011-12-08 Girija Chetty Biometric person identity verification base on face and gait fusion

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
XIAOLI ZHOU等: "Feature fusion of side face and gait for video-based human identification", 《PATTERN RECOGNITION》 *
肖宾杰: "基于图像质量加权的D-S证据理论多生物特征融合识别", 《计算机应用》 *
马樱: "基于视频流的步态识别", 《中国优秀硕士学位论文全文数据库》 *

Cited By (37)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108108693B (en) * 2017-12-20 2019-02-19 深圳市安博臣实业有限公司 Intelligent identification monitoring device and recognition methods based on 3D high definition VR panorama
CN108108693A (en) * 2017-12-20 2018-06-01 深圳市安博臣实业有限公司 Intelligent identification monitoring device and recognition methods based on 3D high definition VR panoramas
CN108460364A (en) * 2018-03-27 2018-08-28 百度在线网络技术(北京)有限公司 Method and apparatus for generating information
CN108460364B (en) * 2018-03-27 2022-03-11 百度在线网络技术(北京)有限公司 Method and apparatus for generating information
CN110443232A (en) * 2018-04-12 2019-11-12 腾讯科技(深圳)有限公司 Method for processing video frequency and relevant apparatus, image processing method and relevant apparatus
CN108629299B (en) * 2018-04-24 2021-11-16 武汉幻视智能科技有限公司 Long-time multi-target tracking method and system combining face matching
CN108734107B (en) * 2018-04-24 2021-11-05 武汉幻视智能科技有限公司 Multi-target tracking method and system based on human face
CN108734107A (en) * 2018-04-24 2018-11-02 武汉幻视智能科技有限公司 A kind of multi-object tracking method and system based on face
CN108629299A (en) * 2018-04-24 2018-10-09 武汉幻视智能科技有限公司 A kind of matched long-time multi-object tracking method of combination face and system
CN109145742A (en) * 2018-07-19 2019-01-04 银河水滴科技(北京)有限公司 A kind of pedestrian recognition method and system
CN109145742B (en) * 2018-07-19 2021-05-11 银河水滴科技(宁波)有限公司 Pedestrian identification method and system
CN109086859A (en) * 2018-08-01 2018-12-25 上海小蚁科技有限公司 People flow rate statistical method, apparatus and computer readable storage medium
CN109325423B (en) * 2018-08-29 2021-09-21 安徽超清科技股份有限公司 SSD (solid State disk) optimization method for pedestrian detection
CN109325423A (en) * 2018-08-29 2019-02-12 安徽超清科技股份有限公司 A kind of optimization SSD algorithm for pedestrian detection
CN109359544A (en) * 2018-09-19 2019-02-19 武汉烽火众智数字技术有限责任公司 A kind of portrait search method and device
CN109359544B (en) * 2018-09-19 2022-01-21 武汉烽火众智数字技术有限责任公司 Portrait retrieval method and device
CN109446936A (en) * 2018-10-12 2019-03-08 银河水滴科技(北京)有限公司 A kind of personal identification method and device for monitoring scene
CN109409297A (en) * 2018-10-30 2019-03-01 咪付(广西)网络技术有限公司 A kind of personal identification method based on binary channels convolutional neural networks
CN109409297B (en) * 2018-10-30 2021-11-23 咪付(广西)网络技术有限公司 Identity recognition method based on dual-channel convolutional neural network
CN109754461A (en) * 2018-12-29 2019-05-14 深圳云天励飞技术有限公司 Image processing method and related product
CN109753917A (en) * 2018-12-29 2019-05-14 中国科学院重庆绿色智能技术研究院 Face quality optimization method, system, computer readable storage medium and equipment
CN109920111A (en) * 2019-03-05 2019-06-21 杭州电子科技大学 A kind of access control system of recognition of face and Gait Recognition fusion
CN110175553A (en) * 2019-05-23 2019-08-27 银河水滴科技(北京)有限公司 The method and device of feature database is established based on Gait Recognition and recognition of face
CN110175587A (en) * 2019-05-30 2019-08-27 黄岩 A kind of video frequency tracking method based on recognition of face and Algorithm for gait recognition
WO2020249025A1 (en) * 2019-06-12 2020-12-17 杭州海康威视数字技术股份有限公司 Identity information determining method and apparatus, and storage medium
CN110276288A (en) * 2019-06-17 2019-09-24 北京影谱科技股份有限公司 A kind of personal identification method and device based on biological characteristic
CN110569707A (en) * 2019-06-25 2019-12-13 深圳和而泰家居在线网络科技有限公司 identity recognition method and electronic equipment
CN110287889A (en) * 2019-06-26 2019-09-27 银河水滴科技(北京)有限公司 A kind of method and device of identification
CN110750671A (en) * 2019-09-05 2020-02-04 浙江省北大信息技术高等研究院 Pedestrian retrieval method and device based on massive unstructured features
CN110796094A (en) * 2019-10-30 2020-02-14 上海商汤智能科技有限公司 Control method and device based on image recognition, electronic equipment and storage medium
CN110874582A (en) * 2019-11-18 2020-03-10 浙江大华技术股份有限公司 Telecommunication fraud determination method and device, storage medium and electronic device
CN111460884A (en) * 2020-02-09 2020-07-28 天津博宜特科技有限公司 Multi-face recognition method based on human body tracking
CN111881732A (en) * 2020-06-17 2020-11-03 北京邮电大学 SVM (support vector machine) -based face quality evaluation method
CN112287770A (en) * 2020-10-10 2021-01-29 武汉大学 Face quality sensing method and system for identity recognition
CN112364829A (en) * 2020-11-30 2021-02-12 北京有竹居网络技术有限公司 Face recognition method, device, equipment and storage medium
CN112488071B (en) * 2020-12-21 2021-10-26 重庆紫光华山智安科技有限公司 Method, device, electronic equipment and storage medium for extracting pedestrian features
CN112488071A (en) * 2020-12-21 2021-03-12 重庆紫光华山智安科技有限公司 Method, device, electronic equipment and storage medium for extracting pedestrian features

Similar Documents

Publication Publication Date Title
CN107590452A (en) A kind of personal identification method and device based on gait and face fusion
CN103942577B (en) Based on the personal identification method for establishing sample database and composite character certainly in video monitoring
CN106295522B (en) A kind of two-stage anti-fraud detection method based on multi-orientation Face and environmental information
Zeng et al. Silhouette-based gait recognition via deterministic learning
CN101558996B (en) Gait recognition method based on orthogonal projection three-dimensional reconstruction of human motion structure
CN101630364A (en) Method for gait information processing and identity identification based on fusion feature
Kusakunniran et al. Automatic gait recognition using weighted binary pattern on video
CN104036236B (en) A kind of face gender identification method based on multiparameter exponential weighting
CN103718175A (en) Apparatus, method, and medium detecting object pose
CN104200200B (en) Fusion depth information and half-tone information realize the system and method for Gait Recognition
CN101807245A (en) Artificial neural network-based multi-source gait feature extraction and identification method
Ahmed et al. Kinect-based gait recognition using sequences of the most relevant joint relative angles
CN106295544B (en) A kind of unchanged view angle gait recognition method based on Kinect
CN105138995B (en) The when constant and constant Human bodys' response method of view based on framework information
CN105740779A (en) Method and device for human face in-vivo detection
CN106599785A (en) Method and device for building human body 3D feature identity information database
CN105320948A (en) Image based gender identification method, apparatus and system
Huang et al. Hand vein recognition based on oriented gradient maps and local feature matching
Patruno et al. People re-identification using skeleton standard posture and color descriptors from RGB-D data
Pandey et al. Hand gesture recognition for sign language recognition: A review
CN102270308A (en) Facial feature location method based on five sense organs related AAM (Active Appearance Model)
CN109522853B (en) Face datection and searching method towards monitor video
Yao et al. Robust gait recognition using hybrid descriptors based on skeleton gait energy image
CN107798308A (en) A kind of face identification method based on short-sighted frequency coaching method
Sun et al. Human recognition for following robots with a Kinect sensor

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20180116