AY. YY TY al
2 TS Om
CS CC. —
Figure 3: Overlay (addition and rescaling) of the two map- Figure 4: Resampling coordinate displacement image for the
registered images of 1991 and 1995 (using thin-plate spline ^ 1995 data between global polynomial and local AKIMA reg-
registration with U(r) — r? Inr). istration (black is vanishing, white is maximum difference,
GCPs marked as white crosses).
VARIANCE OF SECOND PRINCIPAL COMPONENT (‘CHANGE’)
Spectral Band Number i 1 2 3 4 5 6 7 8 9 10
Center Wavelength [um] 0.435 0.485 0.560 0.615 0.660 0.723 0.830 0.980 1.650 2.215
GLOBAL POLYNOMIAL
Absolute Variance of 2. PC: | 26.64 5842 72.58 103.11 113.80 151.74 533.08 503.06 256.99 264.20
Relative Reduction of 2. PC Variance: mean + stdev.:
AKIMA LOCAL POLYNOMIAL | -4.3% -74% -7.0% -6.6% -5.6% —54% -71% -93% -9.2% _43% -(6.6 + 1.8)%
AFFINE + r2In7 | -54% -8.4% 91% 91% -72% -87% -94% -11.9% -120% -6.0% ; -(8.7 + 2.2)%
AFFINE +7 | 49% -85% -7.9% -82% -59% -6.8% -99% -13.5% 12.5% -53% | -(8.3 € 2.9)96
v1+r? |-10% -32% -12% -06% —01% +1.6% -18% -2.2% +2.4% +4.9% | -(0.1 x 2.4)96
POLYNOMIAL + V1 +r? | -52% -93% -9.3% -88% -78% -75% -8.1% -10.1% -10.7% -53% | -(8.2 X 1.8)96
Table 1: The variance of the second principal component is considered as the amount of ‘change’ between the two images of
the scene. Erroneous ‘change’ is reduced by improved registration techniques.
954
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996