Lab 1: Canopy Analysis / Thresholding |
||
|
-- Images -- Example Mask File American Beech, June: Raw Image Manual Threshold: 200 Automatic Threshold: Single Region Automatic Threshold: Multi-Region Norway Maple, June: Raw Image Manual Threshold: 150 Automatic Threshold: Single Region Automatic Threshold: Multi-Region -- Histograms -- American Beech, June Norway Maple, June |
Careful examination of the images produced revealed the weaknesses
of both manual and automatic determination of a proper threshold. Manual
determination tended to set the threshold too high, eliminating areas
which are clearly sky upon human examination of the image. The automatic
thresholding set, in general, a far more intelligent threshold for each
image; but in very low gradient areas(ie, where there is a bright
area of leaves right between a darker area of leaves and a very bright
area of sky), the threshold tends to be lower than it ought to be. The
American Beech images are a good example of where manual thresholding
has great difficulty, and the Norway Maple images show where automatic
thresholding fails partially(look at the brightest region).
Unsurprisingly, automatic multi-region thresholding set more intelligent thresholds than automatic single-region tresholding. We generated circular regions encompassing particular ranges of radii, each about 10 degrees in width, from 0 to 60 degrees(120 degrees in diameter). |
Back |
| Swarthmore College - E/CS 27 - Fall 2003 | ||
