CN109815815B - Pedestrian re-identification method based on metric learning and support vector machine integration - Google Patents

Pedestrian re-identification method based on metric learning and support vector machine integration Download PDF

Info

Publication number
CN109815815B
CN109815815B CN201811576219.0A CN201811576219A CN109815815B CN 109815815 B CN109815815 B CN 109815815B CN 201811576219 A CN201811576219 A CN 201811576219A CN 109815815 B CN109815815 B CN 109815815B
Authority
CN
China
Prior art keywords
pedestrian
support vector
vector machine
identification
metric learning
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
Application number
CN201811576219.0A
Other languages
Chinese (zh)
Other versions
CN109815815A (en
Inventor
李华锋
赵丹丹
王红斌
余正涛
线岩团
文永华
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.)
Kunming University of Science and Technology
Original Assignee
Kunming University of Science and Technology
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 Kunming University of Science and Technology filed Critical Kunming University of Science and Technology
Priority to CN201811576219.0A priority Critical patent/CN109815815B/en
Publication of CN109815815A publication Critical patent/CN109815815A/en
Application granted granted Critical
Publication of CN109815815B publication Critical patent/CN109815815B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Abstract

The invention relates to a pedestrian re-identification method based on metric learning and support vector machine integration, and belongs to the technical field of image processing and pattern recognition. Firstly, generating a pedestrian feature matrix with pedestrian label information; processing a nonlinear space M for measuring the distance of the pedestrian; setting pedestrian label information used in a support vector machine; introducing a constraint variable into the support vector machine, and using the support vector machine as a constraint condition of a nonlinear space; carrying out scaling processing on the constraint condition of the nonlinear space M; and finding the optimal solution of the projection matrix and the classifier, and carrying out pedestrian identification by using an identification model integrating metric learning and a support vector machine to obtain the identification rate. The invention integrates metric learning and a support vector machine. Compared with the prior art, the method provided by the invention effectively excavates and utilizes the tag information in the pedestrian data set, so that the pedestrian matching rate is effectively improved.

Description

Pedestrian re-identification method based on metric learning and support vector machine integration
Technical Field
The invention relates to a pedestrian re-identification method based on metric learning and support vector machine integration, and belongs to the technical field of image processing and pattern recognition.
Background
With the popularization of national smart cities and safe cities, the video monitoring system basically covers the main cities of China. The pedestrian information data stored by the video monitoring system is huge, the efficiency of manual information processing is low, the cost is high, and the pedestrian re-identification technology can effectively improve the working efficiency and save resources. The main task of pedestrian re-identification is to match whether the pedestrian coming from under the non-overlapping cameras is the same pedestrian. The technology is to extract a pedestrian as a target pedestrian from the camera a, judge whether the pedestrian appears under the camera b, and find the target pedestrian if the pedestrian appears. Therefore, the pedestrian re-recognition technology is receiving great attention from a large number of researchers.
At present, although the re-identification of pedestrians with strong robustness draws the attention of researchers and proposes some feasible solutions, the experimental effect still cannot meet the practical needs, especially when the pedestrians intentionally change the appearance of their external features. The existing pedestrian re-identification technology still has the problem that the same pedestrian presents different bottom layer visual characteristics under different vision, different illumination conditions and different cameras due to the change of visual angles and illumination and the difference of the parameter settings of the cameras, and different pedestrians present similar visual characteristics, so that the pedestrian re-identification technology can not meet the requirements of practical application far away.
The pedestrian re-identification research results mainly comprise two categories of pedestrian re-identification based on features and pedestrian re-identification based on metric learning, and the features of the pedestrian based on the features are extracted from the features of the bottom layer of the pedestrian in identification, wherein the features are higher in distinguishing capability and expression capability. The pedestrian re-identification based on metric learning is an algorithm which is better in matching effect and is searched from the perspective of metric learning. The pedestrian re-identification method based on the features utilizes the label information of pedestrians to extract the bottom layer features of the pedestrians, but the pedestrian re-identification method is influenced by factors such as severe illumination change, different camera parameter settings and the like, the bottom layer features of the pedestrians such as color and texture change greatly, and therefore the pedestrian identification accuracy is low. However, the pedestrian re-identification method based on metric learning takes into account the problems of severe illumination change, different camera parameter settings, pedestrian appearance and the like, and utilizes the tag information which appears in pairs between the same pedestrians at different viewing angles to reduce the metric distance between the same pedestrians. However, only the tag information of the same pedestrian at different viewing angles is considered when the pedestrian tag information is used, and the tag information between different pedestrians at different viewing angles is not considered. The method aims at the problems that the pedestrian re-identification algorithm based on metric learning is poor in identification effect and not strong in robustness under the interference factors of complex backgrounds such as clothes, postures and the like of pedestrians.
Disclosure of Invention
The invention provides a pedestrian re-identification method based on metric learning and support vector machine integration, which is used for mining and fully utilizing label information among pedestrians at different visual angles to improve the accuracy of pedestrian identification.
The technical scheme of the invention is as follows: a pedestrian re-identification method based on metric learning and support vector machine integration comprises the steps of firstly generating a pedestrian feature matrix with pedestrian label information; processing a nonlinear space M for measuring the distance of the pedestrian; setting pedestrian label information used in a support vector machine; introducing a constraint variable into the support vector machine, and using the support vector machine as a constraint condition of a nonlinear space; carrying out scaling processing on the constraint condition of the nonlinear space M; and finding the optimal solution of the projection matrix and the classifier, and carrying out pedestrian identification by using an identification model integrating metric learning and a support vector machine to obtain the identification rate.
Further, the pedestrian re-identification method based on the integration of metric learning and support vector machine comprises the following specific steps:
stpe1, projecting all the features of pedestrians under the a and b view angles to the same nonlinear space M ∈ Rm×nInternal, using formula
Figure BDA0001916852160000021
Finding out the most similar but not self-pedestrian under the a-view and the b-view
Figure BDA0001916852160000022
Generating a pedestrian feature matrix x with pedestrian tag informationc；
Stpe2 requires the ith person in each a-view in the non-linear space M
Figure BDA0001916852160000023
And under b viewing angle and
Figure BDA0001916852160000024
most similar pedestrians but not the same person
Figure BDA0001916852160000025
Is smaller than the ith person in the a view angle
Figure BDA0001916852160000026
In b view with itself
Figure BDA0001916852160000027
Measured distance therebetween, i.e.
Figure BDA0001916852160000028
Stpe3, if all pedestrian features are processed according to Stpe2, the projected features meet the conditions, the overfitting condition occurs, so the projected features meet the conditions that the same pedestrian wears greatly differently, the projected features are processed according to Step2, and if different pedestrians wear similarly, the projected features are not processed at all
Figure BDA0001916852160000029
Taking 0; namely, it is
Figure BDA00019168521600000210
Stpe4, the setting mode of pedestrian's label information does: if it is not
Figure BDA0001916852160000031
And
Figure BDA0001916852160000032
if the person is the same person, the tag information y is setijIs 1, if not-1; namely, it is
Figure BDA0001916852160000033
Wherein the content of the first and second substances,
Figure BDA0001916852160000034
representation in support vector machines
Figure BDA0001916852160000035
A measured distance between pedestrian features;
stpe5 and traditional support vector machine introduce a constraint variable xiijIs an inequality
Figure BDA0001916852160000036
Wherein w is a classifier of a support vector machine; and then introducing the constraint condition of the nonlinear space M into a support vector machine, namely:
Figure BDA0001916852160000037
s.t(yij(w(Mxai-Mxbj)+c)＞1-ξij)
stpe6 constraint of the nonlinear space M (y)ij(w(Mxai-Mxbj)+c)＞1-ξij) Carrying out proper scaling processing, relaxing the condition constraint and optimizing the xiijElimination, i.e.
Figure BDA0001916852160000038
Stpe7, finding the optimal solution of a Stpe6 formula, and training and learning to obtain a projection matrix and a classifier w; and then, carrying out pedestrian identification in an identification model integrating metric learning and support vector machine by using a projection matrix and a classifier w to obtain an identification rate s, wherein the identification model integrating metric learning and support vector machine for obtaining the identification rate is as follows:
Figure BDA0001916852160000039
wherein c serves to limit the range of classification similarity values.
The invention has the beneficial effects that:
the invention integrates metric learning and a support vector machine. Compared with the prior art, the method provided by the invention effectively excavates and utilizes the tag information in the pedestrian data set, so that the pedestrian matching rate is effectively improved.
Drawings
FIG. 1 is a block flow diagram of the present invention; wherein y represents a pedestrian label information matrix used in the support vector machine;
FIG. 2 is a comparison experimental result of metric learning model and metric learning model integrated with support vector machine on four data sets, respectively.
Detailed Description
Example 1: as shown in fig. 1-2, a pedestrian re-identification method based on metric learning and support vector machine integration first generates a pedestrian feature matrix with pedestrian label information; processing a nonlinear space M for measuring the distance of the pedestrian; setting pedestrian label information used in a support vector machine; introducing a constraint variable into the support vector machine, and using the support vector machine as a constraint condition of a nonlinear space; carrying out scaling processing on the constraint condition of the nonlinear space M; and finding the optimal solution of the projection matrix and the classifier, and carrying out pedestrian identification by using an identification model integrating metric learning and a support vector machine to obtain the identification rate.
Further, the pedestrian re-identification method based on the integration of metric learning and support vector machine comprises the following specific steps:
stpe1, projecting all the features of pedestrians under the a and b view angles to the same nonlinear space M ∈ Rm×nInternal, using formula
Figure BDA0001916852160000041
Finding out the most similar but not self-pedestrian under the a-view and the b-view
Figure BDA0001916852160000042
Generating a pedestrian feature matrix x with pedestrian tag informationc；
Stpe2 requires the ith person in each a-view in the non-linear space M
Figure BDA0001916852160000043
And under b viewing angle and
Figure BDA0001916852160000044
most similar pedestrians but not the same person
Figure BDA0001916852160000045
Is smaller than the ith person in the a view angle
Figure BDA0001916852160000046
In b view with itself
Figure BDA0001916852160000047
Measured distance therebetween, i.e.
Figure BDA0001916852160000048
Stpe3, if all pedestrian features are processed according to Stpe2, the projected features meet the conditions, the overfitting condition occurs, so the projected features meet the conditions that the same pedestrian wears greatly differently, the projected features are processed according to Step2, and if different pedestrians wear similarly, the projected features are not processed at all
Figure BDA0001916852160000049
Taking 0; namely, it is
Figure BDA00019168521600000410
Stpe4, the setting mode of pedestrian's label information does: if it is not
Figure BDA0001916852160000051
And
Figure BDA0001916852160000052
if the person is the same person, the tag information y is setijIs 1, if not-1; namely, it is
Figure BDA0001916852160000053
Wherein the content of the first and second substances,
Figure BDA0001916852160000054
representation in support vector machines
Figure BDA0001916852160000055
A measured distance between pedestrian features;
stpe5 and traditional support vector machine introduce a constraint variable xiijIs an inequality
Figure BDA0001916852160000056
Wherein w is a classifier of a support vector machine; and then introducing the constraint condition of the nonlinear space M into a support vector machine, namely:
Figure BDA0001916852160000057
s.t(yij(w(Mxai-Mxbj)+c)＞1-ξij)
stpe6 constraint of the nonlinear space M (y)ij(w(Mxai-Mxbj)+c)＞1-ξij) Carrying out proper scaling processing, relaxing the condition constraint and optimizing the xiijElimination, i.e.
Figure BDA0001916852160000058
Stpe7, finding the optimal solution of a Stpe6 formula, and training and learning to obtain a projection matrix and a classifier w; and then, carrying out pedestrian identification in an identification model integrating metric learning and support vector machine by using a projection matrix and a classifier w to obtain an identification rate s, wherein the identification model integrating metric learning and support vector machine for obtaining the identification rate is as follows:
Figure BDA0001916852160000059
wherein c serves to limit the range of classification similarity values.
In order to compare with the existing method, the invention adopts five data sets of VIPER, iLIDS-IVD, CUHK01, PRID2011 and PRID405S to carry out human re-identification experiments, and adopts the average value of ten-fold cross validation as the final result.
The evaluation index is consistent with the comparison method, and a CMC cumulative curve is used as the evaluation index (although the CMC graph is the matching rate of rank1-rank20, the most important value in the application such as actual picture retrieval is rank 1).
To verify the effect of the support vector machine, the proposed metric model is taken as model one (Ours-svm shown in fig. 2) and the metric learning plus support vector machine model is taken as model two (Ours shown in fig. 2). The CMC plots 2 for the two models over four data sets, VIPER, iLIDS-IVD, PRID2011, and PRID 405S. It can be seen that the metric learning and support vector machine before rank5 has obvious effect on improving the pedestrian recognition rate, and gradually becomes stable as the rank value increases.
The results of comparisons on the VIPeR dataset with models of PCCA, LFDA, KISSME, LADF, Mid-filter, ECM, MFA, kLFDA, RD and SR45 are shown in Table 1. Although the CMC graph is the matching rate of rank1-rank20, in practical picture retrieval and other applications, the most important is the value of rank1, and it can be seen from table 1 that the rank1 and rank5 proposed by us are the best.
Table 1: the results of the comparison of the various methods on the VIPER data set list the match rates (%)% for rank1, rank5, rank10 and rank20
Methods Rank1 Rank5 Rank10 Rank20
PCCA 19.3 48.9 64.9 80.3
LFDA 19.7 46.7 62.1 77.0
KISSME 19.6 48.0 62.2 77.0
LADF 29.3 61.0 76.0 86.2
Mid-filter 29.1 52.3 66.0 79.9
ECM 38.2 67.2 78.3 87.9
MFA 32.2 66.0 79.7 90.6
RD 33.3 41.5 78.4 88.5
kLFDA 32.3 65.8 79.7 90.6
SR 32.9 62.0 75.9 89.2
Ours 55.5 67.2 73.2 80.5
On the iLIDS-IDV dataset, SDALF-SS, Color + LBP + DTW, ISR, DVR, DVDL, PHDL + WHOS + STFV3D, Salience + DVR, KISSME were compared and the results are shown in Table 2. It can be seen from table 2 that the recognition rates of rank1 and rank5 of the method proposed by the present invention are the highest.
Table 2: the results of the comparisons of the various methods on the iLIDS-IVD dataset list the match rates (%)% of rank1, rank5, rank10 and rank20
Figure BDA0001916852160000061
Figure BDA0001916852160000071
The results of the experiments on the CUHK01 data set compared with Rcca, ITML, KISSME, Generic metric68, SalMatch, kLFDA, MidFilter, MirrorKMFA, LOMO + LADF, and LOMO + XQDA are shown in Table 3, and it can be seen from Table 3 that the method provided by the present invention has the best recognition effect on rank 1.
Table 3: the results of the comparison of the various methods on the CUHK01 dataset list the match rates (%)% of rank1, rank5, rank10 and rank20
Methods Rank1 Rank5 Rank10 Rank20
Rcca 14.9 32.6 43.8 55.5
ITML 16.0 35.2 45.6 59.8
KISSME 10.3 27.2 37.5 49.7
GenericMetric 20.0 43.6 56.0 69.3
SalMatch 28.5 45.9 55.7 68.0
kLFDA 26.1 49.4 58.4 71.8
MidFilter 34.3 55.1 65.0 74.9
MirrorKMFA 40.4 64.6 75.3 84.1
LOMO+LADF 58.0 83.7 90.5 94.9
LOMO+XQDA 63.2 83.9 90.0 94.4
Ours 64.5 81.9 85 88
The results of comparison of PPLM, RDC, LOMO + LADF, metricenembemble, LOMO + M, XQDA, LOMO + XQDA, DVR, and saience + DVR on the PRID2011 dataset are shown in table 4, except that rank20 is less than ideal, the rank1, rank2, and rank10 of the present invention all identified the best results.
Table 4: comparison of the various methods on the PRID2011 dataset lists the match rates (%)% for rank1, rank5, rank10, and rank20
Methods Rank1 Rank5 Rank10 Rank20
PPLM 15.0 32.0 42.0 54.0
RDC 15.5 38.8 53.2 69.0
LOMO+LADF 16.2 34.0 44.4 59.5
MetricEnsemble 17.9 39.0 50.0 62.0
LOMO+M 15.2 36.1 48.3 60.4
XQDA 24.6 49.3 62.8 76.3
LOMO+XQDA 26.7 49.9 61.9 73.8
DVR 28.9 55.3 65.5 82.8
Salience+DVR 41.7 64.5 77.5 88.8
Ours 72.3 86.8 92 96.7
The results of comparison of ELF, KISSME, EIML, SCNCD, ECM, TSR, MEDVL, KISSME-MGT, LOMO + LADF, KLFDA-MGT, MirrorKMFA, etc. on the PRID450S data set are shown in Table 5, and rank1 of the method of the present invention is the most effective.
Table 5: the results of the comparison of the various methods on the PRID _450S data set list the match rates (%)' for rank1, rank5, rank10, and rank20
Methods Rank1 Rank5 Rank10 Rank20
ELF 30.6 - 73.6 84.2
KISSME 33.0 - 71.0 79.0
EIML 35.0 - 68.0 77
SCNCD 41.5 66.6 75.9 84.4
ECM 41.9 66.3 76.9 84.9
TSR 44.9 71.7 77.5 86.7
MEDVL 45.9 73.0 82.9 91.1
KISSME-MGT 46.1 73.3 83.3 90.7
LOMO+LADF 47.8 74.7 82.8 90.9
KLFDA-MGT 46.1 73.3 83.3 90.7
MirrorKMFA 55.4 79.3 87.8 93.9
Ours 58.4 68.7 73.1 80.2
While the present invention has been described in detail with reference to the embodiments shown in the drawings, the present invention is not limited to the embodiments, and various changes can be made without departing from the spirit of the present invention within the knowledge of those skilled in the art.

Claims (1)

1. A pedestrian re-identification method based on metric learning and support vector machine integration is characterized in that: firstly, generating a pedestrian feature matrix with pedestrian label information; processing a nonlinear space M for measuring the distance of the pedestrian; setting pedestrian label information used in a support vector machine; introducing a constraint variable into the support vector machine, and using the support vector machine as a constraint condition of a nonlinear space; carrying out scaling processing on the constraint condition of the nonlinear space M; finding the optimal solution of a projection matrix and a classifier, and carrying out pedestrian identification by using an identification model integrating metric learning and a support vector machine to obtain an identification rate;
the pedestrian re-identification method based on the integration of metric learning and the support vector machine comprises the following specific steps:
stpe1, projecting all the features of pedestrians under the a and b view angles to the same nonlinear space M ∈ Rm×nInternal, using formula
Figure FDA0003031058060000011
Finding out the most similar but not self-pedestrian under the a-view and the b-view
Figure FDA0003031058060000012
Generating a pedestrian feature matrix x with pedestrian tag informationc；
Stpe2 requires the ith person in each a-view in the non-linear space M
Figure FDA0003031058060000013
And under b viewing angle and
Figure FDA0003031058060000014
most similar pedestrians but not the same person
Figure FDA0003031058060000015
Is smaller than the ith person in the a view angle
Figure FDA0003031058060000016
In b view with itself
Figure FDA0003031058060000017
Measured distance therebetween, i.e.
Figure FDA0003031058060000018
Stpe3, if all pedestrian features are processed according to Stpe2, the projected features meet the conditions, and overfitting occurs, so that the projected features meet the conditions that the same pedestrian wears greatly differently according to Stpe2, and if different pedestrians wear similarly, no processing is performed
Figure FDA0003031058060000019
Taking 0; namely, it is
Figure FDA00030310580600000110
Stpe4, the setting mode of pedestrian's label information does: if it is not
Figure FDA00030310580600000111
And
Figure FDA00030310580600000112
if the person is the same person, the tag information y is setijIs 1, if not-1; namely, it is
Figure FDA0003031058060000021
Wherein the content of the first and second substances,
Figure FDA0003031058060000022
representation in support vector machines
Figure FDA0003031058060000023
A measured distance between pedestrian features;
stpe5 and traditional support vector machine introduce a constraint variable xiijIs an inequality
Figure FDA0003031058060000024
Wherein w is a classifier of a support vector machine; and then introducing the constraint condition of the nonlinear space M into a support vector machine, namely:
Figure FDA0003031058060000025
Figure FDA0003031058060000026
stpe6 constraint of the nonlinear space M (y)ij(w(Mxai-Mxbj)+c)＞1-ξij) Carrying out proper scaling processing, relaxing the condition constraint and optimizing the xiijElimination, i.e.
Figure FDA0003031058060000027
Stpe7, finding the optimal solution of a Stpe6 formula, and training and learning to obtain a projection matrix and a classifier w; and then, carrying out pedestrian identification in an identification model integrating metric learning and support vector machine by using a projection matrix and a classifier w to obtain an identification rate s, wherein the identification model integrating metric learning and support vector machine for obtaining the identification rate is as follows:
Figure FDA0003031058060000028
wherein c serves to limit the range of classification similarity values.
CN201811576219.0A 2018-12-22 2018-12-22 Pedestrian re-identification method based on metric learning and support vector machine integration Active CN109815815B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811576219.0A CN109815815B (en) 2018-12-22 2018-12-22 Pedestrian re-identification method based on metric learning and support vector machine integration

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811576219.0A CN109815815B (en) 2018-12-22 2018-12-22 Pedestrian re-identification method based on metric learning and support vector machine integration

Publications (2)

Publication Number Publication Date
CN109815815A CN109815815A (en) 2019-05-28
CN109815815B true CN109815815B (en) 2021-06-18

Family

ID=66602380

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811576219.0A Active CN109815815B (en) 2018-12-22 2018-12-22 Pedestrian re-identification method based on metric learning and support vector machine integration

Country Status (1)

Country Link
CN (1) CN109815815B (en)

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017212206A1 (en) * 2016-06-06 2017-12-14 Cirrus Logic International Semiconductor Limited Voice user interface

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080101705A1 (en) * 2006-10-31 2008-05-01 Motorola, Inc. System for pattern recognition with q-metrics
KR101972356B1 (en) * 2010-12-21 2019-04-25 한국전자통신연구원 An apparatus and a method for detecting upper body
US10552544B2 (en) * 2016-09-12 2020-02-04 Sriram Chakravarthy Methods and systems of automated assistant implementation and management
US20180173940A1 (en) * 2016-12-19 2018-06-21 Canon Kabushiki Kaisha System and method for matching an object in captured images
CN106778921A (en) * 2017-02-15 2017-05-31 张烜 Personnel based on deep learning encoding model recognition methods again
CN107844752A (en) * 2017-10-20 2018-03-27 常州大学 A kind of recognition methods again of the pedestrian based on block rarefaction representation
CN108345860A (en) * 2018-02-24 2018-07-31 江苏测联空间大数据应用研究中心有限公司 Personnel based on deep learning and learning distance metric recognition methods again
CN108509854B (en) * 2018-03-05 2020-11-17 昆明理工大学 Pedestrian re-identification method based on projection matrix constraint and discriminative dictionary learning

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2017212206A1 (en) * 2016-06-06 2017-12-14 Cirrus Logic International Semiconductor Limited Voice user interface

Also Published As

Publication number Publication date
CN109815815A (en) 2019-05-28

Similar Documents

Publication Publication Date Title
Lavi et al. Survey on deep learning techniques for person re-identification task
CN103295009B (en) Based on the license plate character recognition method of Stroke decomposition
CN104299003A (en) Gait recognition method based on similar rule Gaussian kernel function classifier
CN105701495A (en) Image texture feature extraction method
Ren et al. Parallel RCNN: A deep learning method for people detection using RGB-D images
Wang et al. Lane detection algorithm based on density clustering and RANSAC
CN109815815B (en) Pedestrian re-identification method based on metric learning and support vector machine integration
CN108564020B (en) Micro-gesture recognition method based on panoramic 3D image
CN109784261A (en) Pedestrian's segmentation and recognition methods based on machine vision
Yi et al. Face detection method based on skin color segmentation and facial component localization
Jiang A review of person re-identification
Wang et al. Deep features for person re-identification
Xingbao et al. Pedestrian recognition based on saliency detection and Kalman filter algorithm in aerial video
Ye et al. Real-time TV logo detection based on color and HOG features
Li et al. Real and fake label image classification algorithm based on hog and svm
Lu et al. Unstructured road detection from a single image
Tong et al. A noisy-robust approach for facial expression recognition
Cheng et al. Research on Fast Target Detection And Classification Algorithm for Passive Millimeter Wave Imaging
Zhang et al. Improved BOF Method for Person Re-Identification
Yang et al. Study on image recognition and classification of wood skin defects based on BOW model
CN112069989B (en) Face information acquisition and recognition system and method based on SVD algorithm correction
Li et al. Adaboost Face Detection Based on Improved Covariance Feature.
Yao et al. Scene text extraction based on HSL
Huang et al. A dissimilarity kernel with local features for robust facial recognition
He et al. Research on digital image recognition algorithm based on modular intelligent image recognition

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant