Mini-UAVs for Wilderness Search and Rescue
UAVs can be used to aid in wilderness Search and Rescue. Mini-UAVs are small, lightweight aircraft that
can be deployed easily and are relatively inexpensive. These mUAVs are equipped with downward facing
cameras, GPS units and various telemetry sensors. Video and telemetry data are radioed to a nearby
command station where it is collected and processed. This research focuses primarily on the interpretation and
visualization of data collected on the aircraft.
3D Terrain Models Representing the search Scene
To give proper context to the information gathered from a UAV, a 3D terrain model can
be created representing the area to be searched. The Unites States Geological Survey (USGS)
provides Digital Elevation Models (DEMs) of most of the Unites States. These models are
essential height fields and can be downloaded automatically and processed to create
3D polygonal models that can be rendered in DirectX or OpenGL. This terrain model
can be textured with satellite imagery or some other ortho-imagery.
Projecting Video to the Terrain Model
Once a 3D scene has been constructed, the mUAV can be placed in the scene based
on the telemetry data collected from the craft. From this location in the scene, the
video images filmed from the mUAV can be projected down onto the terrain model.
Correcting mUAV Pose using Image Matching and Pose Estimation
Telemetry data is inaccurate and is often not well synchronized with video data. We can correct this
by performing visual matching between a video frame and the virtual video frame filmed of
the terrain model from the pose of the mUAV. Once a correspondence between these two frames is
established, it is possible to estimate the pose of the mUAV with more accuracy. The POSIT algorithm is
one method of estimating pose.
OpenCV
OpenCV is an open source computer vision library developed in part by Intel. The
library contains functions to find interesting points to track in an image, find the
corresponding points in another image using optical flow, and compute camera parameters
from 2D and corresponding 3D points. In addition, the library has an implementation of the
POSIT algorithm.
How can I learn more?
Our UAV research is ongoing. If you would like source code, sample data or just more info,
please contact us at:
product@enghlabs.com
Copyright © 2007 EnghLabs.com