Finding people by their shadows: Aerial surveillance using body biometrics extracted from ground video

Yumi Iwashita, Adrian Stoica, Ryo Kurazume

Research output: Chapter in Book/Report/Conference proceedingConference contribution

4 Citations (Scopus)

Abstract

Shadow analysis has been shown to enable the extension of gait biometrics to aerial surveillance. In past work the classifiers were both trained and tested on shadow features extracted by image processing. In real scenarios this requires imagery with shadows of people to be recognized. On the other hand one rarely has available the shadow information of the person sought, however direct body movement/information may be more easily obtained from ground surveillance cameras or video recordings. This paper proposes a scenario in which gait/dynamics features from body movement are obtained from a ground video and the search for matching dynamics of shadows takes place in aerial surveillance video. A common scenario would be the recording of people by ground/city surveillance cameras and the use of information to initiate a wide-area search for shadows from aerial platforms. Vice-versa, the shadow of a suspect leaving an incident area, detected by aerial surveillance, can trigger a city-wide search on body/gait biometrics as observed with city/ground surveillance cameras. To illustrate the feasibility of this approach the paper introduces a method that compares contours of bodies in ground image frames and contours of shadows in aerial image frames, for which an alignment is made and a distance is calculated, integrated over a normalized gait cycle. While the results are preliminary, for only 5 people, and using a specific walking arrangement to avoid compensation for changes in the viewing angles, the method obtains a 70% correct classification rate which is a first step in proving the feasibility of the approach.

Original languageEnglish
Title of host publicationProceedings - 3rd International Conference on Emerging Security Technologies, EST 2012
Pages43-48
Number of pages6
DOIs
Publication statusPublished - Nov 26 2012
Event3rd International Conference on Emerging Security Technologies, EST 2012 - Lisbon, Portugal
Duration: Sep 5 2012Sep 7 2012

Publication series

NameProceedings - 3rd International Conference on Emerging Security Technologies, EST 2012

Other

Other3rd International Conference on Emerging Security Technologies, EST 2012
CountryPortugal
CityLisbon
Period9/5/129/7/12

Fingerprint

Antenna grounds
Biometrics
Antennas
Cameras
Video recording
Image processing
Classifiers

All Science Journal Classification (ASJC) codes

  • Safety, Risk, Reliability and Quality

Cite this

Iwashita, Y., Stoica, A., & Kurazume, R. (2012). Finding people by their shadows: Aerial surveillance using body biometrics extracted from ground video. In Proceedings - 3rd International Conference on Emerging Security Technologies, EST 2012 (pp. 43-48). [6328079] (Proceedings - 3rd International Conference on Emerging Security Technologies, EST 2012). https://doi.org/10.1109/EST.2012.41

Finding people by their shadows : Aerial surveillance using body biometrics extracted from ground video. / Iwashita, Yumi; Stoica, Adrian; Kurazume, Ryo.

Proceedings - 3rd International Conference on Emerging Security Technologies, EST 2012. 2012. p. 43-48 6328079 (Proceedings - 3rd International Conference on Emerging Security Technologies, EST 2012).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Iwashita, Y, Stoica, A & Kurazume, R 2012, Finding people by their shadows: Aerial surveillance using body biometrics extracted from ground video. in Proceedings - 3rd International Conference on Emerging Security Technologies, EST 2012., 6328079, Proceedings - 3rd International Conference on Emerging Security Technologies, EST 2012, pp. 43-48, 3rd International Conference on Emerging Security Technologies, EST 2012, Lisbon, Portugal, 9/5/12. https://doi.org/10.1109/EST.2012.41
Iwashita Y, Stoica A, Kurazume R. Finding people by their shadows: Aerial surveillance using body biometrics extracted from ground video. In Proceedings - 3rd International Conference on Emerging Security Technologies, EST 2012. 2012. p. 43-48. 6328079. (Proceedings - 3rd International Conference on Emerging Security Technologies, EST 2012). https://doi.org/10.1109/EST.2012.41
Iwashita, Yumi ; Stoica, Adrian ; Kurazume, Ryo. / Finding people by their shadows : Aerial surveillance using body biometrics extracted from ground video. Proceedings - 3rd International Conference on Emerging Security Technologies, EST 2012. 2012. pp. 43-48 (Proceedings - 3rd International Conference on Emerging Security Technologies, EST 2012).
@inproceedings{3e6c4866330d420a90b2453abd39798a,
title = "Finding people by their shadows: Aerial surveillance using body biometrics extracted from ground video",
abstract = "Shadow analysis has been shown to enable the extension of gait biometrics to aerial surveillance. In past work the classifiers were both trained and tested on shadow features extracted by image processing. In real scenarios this requires imagery with shadows of people to be recognized. On the other hand one rarely has available the shadow information of the person sought, however direct body movement/information may be more easily obtained from ground surveillance cameras or video recordings. This paper proposes a scenario in which gait/dynamics features from body movement are obtained from a ground video and the search for matching dynamics of shadows takes place in aerial surveillance video. A common scenario would be the recording of people by ground/city surveillance cameras and the use of information to initiate a wide-area search for shadows from aerial platforms. Vice-versa, the shadow of a suspect leaving an incident area, detected by aerial surveillance, can trigger a city-wide search on body/gait biometrics as observed with city/ground surveillance cameras. To illustrate the feasibility of this approach the paper introduces a method that compares contours of bodies in ground image frames and contours of shadows in aerial image frames, for which an alignment is made and a distance is calculated, integrated over a normalized gait cycle. While the results are preliminary, for only 5 people, and using a specific walking arrangement to avoid compensation for changes in the viewing angles, the method obtains a 70{\%} correct classification rate which is a first step in proving the feasibility of the approach.",
author = "Yumi Iwashita and Adrian Stoica and Ryo Kurazume",
year = "2012",
month = "11",
day = "26",
doi = "10.1109/EST.2012.41",
language = "English",
isbn = "9780769547916",
series = "Proceedings - 3rd International Conference on Emerging Security Technologies, EST 2012",
pages = "43--48",
booktitle = "Proceedings - 3rd International Conference on Emerging Security Technologies, EST 2012",

}

TY - GEN

T1 - Finding people by their shadows

T2 - Aerial surveillance using body biometrics extracted from ground video

AU - Iwashita, Yumi

AU - Stoica, Adrian

AU - Kurazume, Ryo

PY - 2012/11/26

Y1 - 2012/11/26

N2 - Shadow analysis has been shown to enable the extension of gait biometrics to aerial surveillance. In past work the classifiers were both trained and tested on shadow features extracted by image processing. In real scenarios this requires imagery with shadows of people to be recognized. On the other hand one rarely has available the shadow information of the person sought, however direct body movement/information may be more easily obtained from ground surveillance cameras or video recordings. This paper proposes a scenario in which gait/dynamics features from body movement are obtained from a ground video and the search for matching dynamics of shadows takes place in aerial surveillance video. A common scenario would be the recording of people by ground/city surveillance cameras and the use of information to initiate a wide-area search for shadows from aerial platforms. Vice-versa, the shadow of a suspect leaving an incident area, detected by aerial surveillance, can trigger a city-wide search on body/gait biometrics as observed with city/ground surveillance cameras. To illustrate the feasibility of this approach the paper introduces a method that compares contours of bodies in ground image frames and contours of shadows in aerial image frames, for which an alignment is made and a distance is calculated, integrated over a normalized gait cycle. While the results are preliminary, for only 5 people, and using a specific walking arrangement to avoid compensation for changes in the viewing angles, the method obtains a 70% correct classification rate which is a first step in proving the feasibility of the approach.

AB - Shadow analysis has been shown to enable the extension of gait biometrics to aerial surveillance. In past work the classifiers were both trained and tested on shadow features extracted by image processing. In real scenarios this requires imagery with shadows of people to be recognized. On the other hand one rarely has available the shadow information of the person sought, however direct body movement/information may be more easily obtained from ground surveillance cameras or video recordings. This paper proposes a scenario in which gait/dynamics features from body movement are obtained from a ground video and the search for matching dynamics of shadows takes place in aerial surveillance video. A common scenario would be the recording of people by ground/city surveillance cameras and the use of information to initiate a wide-area search for shadows from aerial platforms. Vice-versa, the shadow of a suspect leaving an incident area, detected by aerial surveillance, can trigger a city-wide search on body/gait biometrics as observed with city/ground surveillance cameras. To illustrate the feasibility of this approach the paper introduces a method that compares contours of bodies in ground image frames and contours of shadows in aerial image frames, for which an alignment is made and a distance is calculated, integrated over a normalized gait cycle. While the results are preliminary, for only 5 people, and using a specific walking arrangement to avoid compensation for changes in the viewing angles, the method obtains a 70% correct classification rate which is a first step in proving the feasibility of the approach.

UR - http://www.scopus.com/inward/record.url?scp=84869390844&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84869390844&partnerID=8YFLogxK

U2 - 10.1109/EST.2012.41

DO - 10.1109/EST.2012.41

M3 - Conference contribution

AN - SCOPUS:84869390844

SN - 9780769547916

T3 - Proceedings - 3rd International Conference on Emerging Security Technologies, EST 2012

SP - 43

EP - 48

BT - Proceedings - 3rd International Conference on Emerging Security Technologies, EST 2012

ER -