24
ERDC/CRREL TR-02-11
Study conclusions
The conclusions from this study are that:
1. It was demonstrated that a 53-frame data set can be acquired successfully
in one day and post-processed, including orthorectification and creation
of geoTIFF files, within 4 hr of the plane's landing.
2. It is possible to distinguish individual roof rafters at 8-in. GSD and
plywood-covered roofs from roofs covered with shingles and/or tarpaper
at 2-ft GSD or finer. Shingles and tarpaper are not visually distinct from
each other in these data sets.
3. For high-resolution images (2-ft or finer), a medium angle (sun with
respect to zenith) increases the roof detail.
4. 8-in. GSD is not sufficient to visually distinguish broad-leaf from
narrow-leaf trees, but it does permit visual identification of palms.
Future recommendations
The following recommendations are made with reference to future acquisi-
tions of emergency management imagery and processing:
1. Adjust all electronic camera sensors to avoid saturation of bright objects
in any band because that results in permanent image contrast loss. This
will increase the ability to detect roof rafters on damaged roofs.
2. Fly lower-resolution (3-ft) data with nearly overhead sun illumination
and the higher-resolution (2-ft or finer) data at a medium angle between
sun and zenith, since roof shadows increase the interpretable roof detail
by increasing contrast.
3. Develop an algorithm to count damaged roofs automatically based upon
scatter plots of band 1 vs. band 2 that have shown a linear cluster for
intact roofs and have shown a nondirectional bright-end cluster for
damaged roofs.