Abstract

In person re-identification, it is of great importance toextract very discriminative features in order to distinguishbetween images of different persons captured in differentcamera views. Features can be extracted globally from theentire or parts of the image, or locally from patches. Inaddition, accuracy can be increased by utilizing supervisedlearning before matching. Matching can either be done us-ing global feature descriptors or local patch features withthe latter being more computationally heavy due to the num-ber of patch pairs to match. Horizontal constraints aretherefore usually added to only match patches at same hor-izontal location. As an extension, we have developed an al-gorithm to add vertical constraints to different body parts,to increase accuracy and decrease processing time. Theconstraints are applied to the CVPDL system by Liet al.[13] in order to compare accuracy with the use of a fore-ground mask and processing time when only adding hori-zontal constraints. We refer to our constrained CVPDL asC-CVPDL. Experiments conducted on two datasets, VIPeRand CUHK01, show C-CVPDL to achieve similar rank-1accuracy on VIPeR while improving rank-1 accuracy by3.83% for CUHK01 compared to CVPDL. Furthermore,experimental results on CUHK03 show a rank-1 accuracy52.05%, being comparable to state-of-the-art CNN’s andbeating other patch matching systems. Finally, timings forVIPeR and CUHK01 show our constraints to decrease timeby 32.77% and 37%, respectively, while only taking up to36ms to compute per person.
Original languageEnglish
Title of host publicationIEEE International Conference on Identity, Security and Behavior Analysis (ISBA), 2017
Number of pages8
PublisherIEEE
Publication dateJun 2017
ISBN (Electronic)978-1-5090-5592-0
DOIs
Publication statusPublished - Jun 2017
EventIEEE International Conference on Identity, Security and Behavior Analysis - New Delhi, India
Duration: 22 Feb 201724 Feb 2017

Conference

ConferenceIEEE International Conference on Identity, Security and Behavior Analysis
CountryIndia
CityNew Delhi
Period22/02/201724/02/2017

Fingerprint

Processing
Masks
Experiments

Keywords

  • re-identification
  • vertical constraints
  • biometrics

Cite this

Lejbølle, A. R., Nasrollahi, K., & Moeslund, T. B. (2017). Constraint Patch Matching for Faster Person Re-identification. In IEEE International Conference on Identity, Security and Behavior Analysis (ISBA), 2017 IEEE. https://doi.org/10.1109/ISBA.2017.7947703
Lejbølle, Aske Rasch ; Nasrollahi, Kamal ; Moeslund, Thomas B. / Constraint Patch Matching for Faster Person Re-identification. IEEE International Conference on Identity, Security and Behavior Analysis (ISBA), 2017 . IEEE, 2017.
@inproceedings{ab73ca48e4be4751bfc25196cac1493e,
title = "Constraint Patch Matching for Faster Person Re-identification",
abstract = "In person re-identification, it is of great importance toextract very discriminative features in order to distinguishbetween images of different persons captured in differentcamera views. Features can be extracted globally from theentire or parts of the image, or locally from patches. Inaddition, accuracy can be increased by utilizing supervisedlearning before matching. Matching can either be done us-ing global feature descriptors or local patch features withthe latter being more computationally heavy due to the num-ber of patch pairs to match. Horizontal constraints aretherefore usually added to only match patches at same hor-izontal location. As an extension, we have developed an al-gorithm to add vertical constraints to different body parts,to increase accuracy and decrease processing time. Theconstraints are applied to the CVPDL system by Liet al.[13] in order to compare accuracy with the use of a fore-ground mask and processing time when only adding hori-zontal constraints. We refer to our constrained CVPDL asC-CVPDL. Experiments conducted on two datasets, VIPeRand CUHK01, show C-CVPDL to achieve similar rank-1accuracy on VIPeR while improving rank-1 accuracy by3.83{\%} for CUHK01 compared to CVPDL. Furthermore,experimental results on CUHK03 show a rank-1 accuracy52.05{\%}, being comparable to state-of-the-art CNN’s andbeating other patch matching systems. Finally, timings forVIPeR and CUHK01 show our constraints to decrease timeby 32.77{\%} and 37{\%}, respectively, while only taking up to36ms to compute per person.",
keywords = "re-identification, vertical constraints, biometrics",
author = "Lejb{\o}lle, {Aske Rasch} and Kamal Nasrollahi and Moeslund, {Thomas B.}",
year = "2017",
month = "6",
doi = "10.1109/ISBA.2017.7947703",
language = "English",
booktitle = "IEEE International Conference on Identity, Security and Behavior Analysis (ISBA), 2017",
publisher = "IEEE",
address = "United States",

}

Lejbølle, AR, Nasrollahi, K & Moeslund, TB 2017, Constraint Patch Matching for Faster Person Re-identification. in IEEE International Conference on Identity, Security and Behavior Analysis (ISBA), 2017 . IEEE, IEEE International Conference on Identity, Security and Behavior Analysis, New Delhi, India, 22/02/2017. https://doi.org/10.1109/ISBA.2017.7947703

Constraint Patch Matching for Faster Person Re-identification. / Lejbølle, Aske Rasch; Nasrollahi, Kamal; Moeslund, Thomas B.

IEEE International Conference on Identity, Security and Behavior Analysis (ISBA), 2017 . IEEE, 2017.

Research output: Contribution to book/anthology/report/conference proceedingArticle in proceedingResearchpeer-review

TY - GEN

T1 - Constraint Patch Matching for Faster Person Re-identification

AU - Lejbølle, Aske Rasch

AU - Nasrollahi, Kamal

AU - Moeslund, Thomas B.

PY - 2017/6

Y1 - 2017/6

N2 - In person re-identification, it is of great importance toextract very discriminative features in order to distinguishbetween images of different persons captured in differentcamera views. Features can be extracted globally from theentire or parts of the image, or locally from patches. Inaddition, accuracy can be increased by utilizing supervisedlearning before matching. Matching can either be done us-ing global feature descriptors or local patch features withthe latter being more computationally heavy due to the num-ber of patch pairs to match. Horizontal constraints aretherefore usually added to only match patches at same hor-izontal location. As an extension, we have developed an al-gorithm to add vertical constraints to different body parts,to increase accuracy and decrease processing time. Theconstraints are applied to the CVPDL system by Liet al.[13] in order to compare accuracy with the use of a fore-ground mask and processing time when only adding hori-zontal constraints. We refer to our constrained CVPDL asC-CVPDL. Experiments conducted on two datasets, VIPeRand CUHK01, show C-CVPDL to achieve similar rank-1accuracy on VIPeR while improving rank-1 accuracy by3.83% for CUHK01 compared to CVPDL. Furthermore,experimental results on CUHK03 show a rank-1 accuracy52.05%, being comparable to state-of-the-art CNN’s andbeating other patch matching systems. Finally, timings forVIPeR and CUHK01 show our constraints to decrease timeby 32.77% and 37%, respectively, while only taking up to36ms to compute per person.

AB - In person re-identification, it is of great importance toextract very discriminative features in order to distinguishbetween images of different persons captured in differentcamera views. Features can be extracted globally from theentire or parts of the image, or locally from patches. Inaddition, accuracy can be increased by utilizing supervisedlearning before matching. Matching can either be done us-ing global feature descriptors or local patch features withthe latter being more computationally heavy due to the num-ber of patch pairs to match. Horizontal constraints aretherefore usually added to only match patches at same hor-izontal location. As an extension, we have developed an al-gorithm to add vertical constraints to different body parts,to increase accuracy and decrease processing time. Theconstraints are applied to the CVPDL system by Liet al.[13] in order to compare accuracy with the use of a fore-ground mask and processing time when only adding hori-zontal constraints. We refer to our constrained CVPDL asC-CVPDL. Experiments conducted on two datasets, VIPeRand CUHK01, show C-CVPDL to achieve similar rank-1accuracy on VIPeR while improving rank-1 accuracy by3.83% for CUHK01 compared to CVPDL. Furthermore,experimental results on CUHK03 show a rank-1 accuracy52.05%, being comparable to state-of-the-art CNN’s andbeating other patch matching systems. Finally, timings forVIPeR and CUHK01 show our constraints to decrease timeby 32.77% and 37%, respectively, while only taking up to36ms to compute per person.

KW - re-identification

KW - vertical constraints

KW - biometrics

U2 - 10.1109/ISBA.2017.7947703

DO - 10.1109/ISBA.2017.7947703

M3 - Article in proceeding

BT - IEEE International Conference on Identity, Security and Behavior Analysis (ISBA), 2017

PB - IEEE

ER -

Lejbølle AR, Nasrollahi K, Moeslund TB. Constraint Patch Matching for Faster Person Re-identification. In IEEE International Conference on Identity, Security and Behavior Analysis (ISBA), 2017 . IEEE. 2017 https://doi.org/10.1109/ISBA.2017.7947703