Detail of the student project

List
Topic:Monocular body height measurement
Department:Katedra kybernetiky
Supervisor:Ing. Jan Čech Ph.D.
Announce as:DP,BP,SOP,PRO
Description:single view metrology
A human body height is a cue which could be used in identity recognition in a surveillance setup. Identity of a subject is usually recognized from face images, which is however not always available in a sufficient quality, due to a limited resolution, distance, or occlusion. Therefore multiple cues is fused in order to recognize the subject's identity more reliably: face recognition, body height, or gait [1,2].

Body height can be measured absolutely with a single camera in a calibrated setup [3], or at least relatively (e.g. who is taller than who/what) if the calibration information is limited or unavailable.

This topic would require designing a ground-truth experiment in order to study various aspects as camera resolution, camera angle, subject distance, and their impact on accuracy of the measurement. A fusion with face recognition could be also further investigated.

The image is a reproduction from [3].

References

[1] Worapan Kusakunniran, Qiang Wu, Jian Zhang, Yi Ma, and Hongdong Li. A New View-Invariant Feature for Cross-View Gait Recognition. IEEE Trans. on Information Forensics and Security, 8(10), 2012.
[2] Sudeep Sarkar, P. Jonathon Phillips, Zongyi Liu, Isidro Robledo Vega, Patrick Grother, and Kevin W. Bowyer. The HumanID Gait Challenge Problem: Data Sets, Performance, and Analysis. IEEE Trans. on PAMI, 27(2), 2005.
[3] A. Criminisi, I. Reid and A. Zisserman. Single View Metrology. International Journal of Computer Vision 40(2), 2000.
Realization form:software, technical report
Date:13.05.2014
Responsible person: Petr Pošík