CONCLUSIONS - TR02-110031

Evaluation of New Sensors for Emergency Management

23

6

CONCLUSIONS

Detectable-objects matrix

Table 1 shows a detection matrix for objects as a function of GSD. Individual

roof rafters are always distinguishable with 8-in. GSD and often with 1-ft GSD.

Wood covering vs. tarpaper and/or shingles can always be determined with 1-ft

and 8-in. GSD, and often with 2-ft GSD. Saturation of all three bands and the

resulting absence of contrast tend to obscure rafter detail.

Information in cloud shadows is not completely recoverable, but if there are

few of them, sufficient solar illumination, and enough sensor quantization levels,

then some information is available from the shadowed regions. Palms can be dis-

tinguished from other species with 8-in. GSD. The best display for visual inter-

pretation of objects with these data sets was bands 1, 2, and 3 in red, green, and

blue with a 2Σ (standard-deviation) stretch to enhance image detail. We used the

ERDAS image processing program (ERDAS 1999).

Table 1. Detection matrix for objects at various GSDs.

Objects/GSD

3-ft

2-ft

1-ft

8-in.

Roof rafters

Not visible

Barely visible

Often visible

Visible

Shingles/tar-

Can determine

Can sometimes

Can often

paper (other)

wood vs. other

separate

vs. plywood

cover

Rafters in 3-

Causes rafter

band satura-

detail loss

tion

Broad-leaf vs.

Cannot

Can determine

Palms are

narrow-leaf

separate

leaf on/off

always visible

All in cloud

Some info

Degrades image

shadow

recoverable

Roofs as a

Best detail, near

Best detail,

function of sun

zero angle,

medium angle,

to zenith angle

overhead sun

shadow casting

All in1,2,3

Enhances

RGB, 2Σ

imagery

stretch