—
Meters
ors
go
€
2
©
2
>
w
504
N d ^ P J T
Vv T N N a4 N ^N f, photo
P Xu b^ El Cer" M v^] (Nw Na N vu
\ | wv v
wr | i 7 M \
M | Ld xd LA gum
ov x us SEE ren iD mE, ur NA xd ud Y rade
photo
50+
-100
100 200 300 400 500 600 700 800 900 1000
Horizonta! distance (m)
FIGURE 6. Ground and tree canopy profiles generated photogrammetrically
and by the radar altimeter. Vertical lines represent spot checks
of tree height.
during the radar altimeter tests showed particular promise to the local scientific
personnel and prompted us to examine their potential. To do so, the following limited
test was conducted.
A map of part of one of the 200 by 200 m plots was available with every tree
mapped in position and containing species codes and diameter information. Seventy-
three of these trees representing 12 species could positively be identified on the large-
scale photos. One of each species was selected for a preliminary test to act as a
standard for training, reference and coding purposes and to screen the following photo
coverage: 70 mm black and white 4X enlargements (scale about 1:500), 70 mm colour
contact prints (scale about 1:2000), 230 mm black and white contact prints (scale about
1:2700) and 230 mm colour contact prints (scale about 1:2400). The remaining trees were
withheld for evaluating the accuracy of the species identification.
The main test was carried out using the 4X black and white enlargements
agreed upon as "best" by the three interpreters. The interpretation results were plotted
in the commonly used "confusion matrix" to analyze the species identification accuracy.
Results
There remained no doubt that the altimeter's signal did penetrate the foliage
and that the profiles obtained were indeed those of the terrain. After the first flights
over line 1 and comparisons of radar traces with the surveyed profiles, a significant and
consistent discrepancy was noted. A field check of the surveyed line revealed a 4.2 m
error in the field survey. After correction, the profiles matched very well (Figure 3).
The extreme deviations are in the order of 5 m, the standard deviation of the differences