-
Home
- Fall 2009 elec301 group project
- Fall 2009 elec301 group project
- Introduction
Introduction
A common problem in dealing with Unmanned Aerial Vehicles
(UAVs) is image stabilization. If an operator wishes to controlthe craft in real-time, a camera mounted on the UAV is often a good solution. This video feed, if left in its original state,
has varying amounts of jitter, which in turn makes operatingthe craft more difficult and makes the footage of the flight much less pleasant to watch.
We decided that we could stabilize the video without using anyadditional hardware-based assistance (such as gyroscopes)
with the digital signal processing techniques we've learned overthe semester.
Our first approach to solving this problem was to correlate eachvideo frame with the previous one, but this proved to be less than optimal ; there exists a faster, more accurate technique. Enter KLT
feature tracking and Serial Affine Transformation.We used a freely-available KLT feature tracker from Stan Birchfield,
then prototyped our affine transformation techniques in MATLAB.We have started porting our work to C, and in the future we expect
this sort of solution to be fully implemented on GPUs for real-timeprocessing.
Source:
OpenStax, Fall 2009 elec301 group project report: video stabilization. OpenStax CNX. Dec 21, 2009 Download for free at http://cnx.org/content/col11152/1.1
Google Play and the Google Play logo are trademarks of Google Inc.
