Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Change detection on UGV patrols with respect to a reference tour using VIS imagery

: Müller, Thomas

Postprint urn:nbn:de:0011-n-3526322 (712 KByte PDF)
MD5 Fingerprint: a85be1b9d84acbb81be371ae8b246c3d
Copyright Society of Photo-Optical Instrumentation Engineers. One print or electronic copy may be made for personal use only. Systematic reproduction and distribution, duplication of any material in this paper for a fee or for commercial purposes, or modification of the content of the paper are prohibited.
Created on: 6.8.2015

Henry, Daniel J. (Ed.) ; Society of Photo-Optical Instrumentation Engineers -SPIE-, Bellingham/Wash.; Society for Imaging Science and Technology -IS&T-:
Airborne Intelligence, Surveillance, Reconnaissance (ISR) Systems and Applications XII : 20-21 April 2015, Baltimore, Maryland, United States
Bellingham, WA: SPIE, 2015 (Proceedings of SPIE 9460)
ISBN: 9781628415766
Paper 94600I, 12 pp.
Conference "Airborne Intelligence, Surveillance, Reconnaissance (ISR) Systems and Applications" <12, 2015, Baltimore/Md.>
Conference Paper, Electronic Publication
Fraunhofer IOSB ()
change detection; detection of dangerous situations; image retrieval; image registration; 3-D parallax; Unmanned Ground Vehicle (UGV); Counter-IED; security and safety

Autonomous driving robots (UGVs, Unmanned Ground Vehicles) equipped with visual-optical (VIS) cameras offer a high potential to automatically detect suspicious occurrences and dangerous or threatening situations on patrol. In order to explore this potential, the scene of interest is recorded first on a reference tour representing the 'everything okay' situation. On further patrols changes are detected with respect to the reference in a two step processing scheme. In the first step, an image retrieval is done to find the reference images that are closest to the current camera image on patrol. This is done efficiently based on precalculated image-to-image registrations of the reference by optimizing image overlap in a local reference search (after a global search when that is needed). In the second step, a robust spatio-temporal change detection is performed that widely compensates 3-D parallax according to variations of the camera position. Various results document the performance of the presented approach.