*** Apologies for Cross-posting *** Toby Breckon, Applied Mathematics and Computing Group, School of Engineering, Cranfield University "Automatic Object Detection and Recognition from UAV Platforms"
Friday 16th October 2009, 2-3pm Room 1.10, Kilburn Building, The University of Manchester Abstract: UAV (Unmanned Aerial Vehicle) platforms are increasingly being considered for search and surveillance operations both on land and at sea due to factors such as cost, convenience and reduced risks. However, the use of UAV sensor platforms in such scenarios has a significant overhead of the manual analysis of extensive aerial video imagery. Often this imagery consists of vast areas of barren seascape or farm/dessert lands with the seldom appearance of objects of interest to the operator. The manual screening of such imagery is both resource consuming and subject to human error. This issue scales with the number of such platforms deployed in any given search or surveillance operation. Here we discuss recent work in the automatic detection and recognition of vehicles, people and generic salient objects (e.g. crash wreckage) from UAV platforms. Work within this area is limited and we discuss the many challenges of working with UAV imagery and the additional challenges of experimental work with UAV platforms. Several research results from our work in this domain are presented including that from the MoD Grand Challenge 2008 which formed part of the Stellar Team's winning SATURN system. Part of http://manchester.siggraph.org/ ----- Anyone wishing to view a discussion via Access Grid should note the following. Please check the Jabber room shortly before the start of the seminar to find if ScreenStreamer is to be used. Virtual Venue: University of Manchester Room1:10 Jabber Room: university-of-manchester-(1.10) (sam.ag.manchester.ac.uk)@conference.mcs.anl.gov ScreenStreamer: Preferably use ScreenStreamer software - go to http://www.memetic-vre.net/software/ScreenStreamer/ScreenStreamer.php and select room "University of Manchester (1.10)" from the AGSC section. Grab the screen feed (by clicking on it) from the screenstreamer window and resize it - more info can be found on http://www.memetic-vre.net/software/ScreenStreamer. Alternatively, the screenshot (not animated) can be accessed at http://192.150.184.66:7890 (this IP might change for some sessions - please check Jabber).
