Testing the method when reference bands are not congruous

Because some airborne systems may have spectral bands that may not be covered by congruous spectral bands on a satellite system, it was important to test surfaces developed where the reference bands do not overlap the spectral band that is to be corrected.

This has been tested with Daedalus data over tropical woodlands, a mine site, wetlands and a sandstone escarpments in the NT; and with DMSV over woodlands and exposed geology in the Pilbara, WA. The modelled surfaces were shown in Figure 7 and Figure 9. Some bands are slightly offset from the Landsat TM bands. In both cases the illumination effects were successfully removed. The following example is a test of bands that do not overlap with Landsat TM using Daedalus bands 4, 6 and 8. These were modelled with bands 1, 2, 3, 4 and 5 of Landsat TM. Figure 15 shows a mosaic of the Daedalus raw images using bands 4, 6 and 8 on the left and a mosaic of the corrected data using the same bands.

Correlation	T1(450-520)	T2(520-600)	T3(630-690)	T4(760-900)	T5(1550-1750)	T7(2080-2350)	D1(420-450)	D2(450-520)	D3(520-600)	D4(605-625)	D5(630-690)	D6(695-750)	D7(760-900)	D8(910-1050)	D9(1550-1750)	D10(2080-2350)
T1(450-520)	1	0.95	0.912	0.486	0.65	0.702	0.669	0.653	0.639	0.665	0.585	0.453	0.368	0.397	0.483	0.461
T2(520-600)	0.95	1	0.96	0.608	0.709	0.712	0.652	0.659	0.673	0.7	0.636	0.526	0.451	0.475	0.526	0.469
T3(630-690)	0.912	0.96	1	0.629	0.791	0.766	0.626	0.664	0.692	0.728	0.688	0.551	0.479	0.508	0.594	0.519
T4(760-900)	0.486	0.608	0.629	1	0.586	0.349	0.331	0.366	0.441	0.46	0.497	0.664	0.71	0.682	0.448	0.247
T5(1550-1750)	0.65	0.709	0.791	0.586	1	0.903	0.406	0.503	0.537	0.559	0.581	0.445	0.429	0.498	0.693	0.582
T7(2080-2350)	0.702	0.712	0.766	0.349	0.903	1	0.442	0.53	0.525	0.538	0.529	0.308	0.25	0.319	0.611	0.628
D1(420-450)	0.669	0.652	0.626	0.331	0.406	0.442	1	0.839	0.814	0.818	0.75	0.621	0.496	0.422	0.582	0.633
D2(450-520)	0.653	0.659	0.664	0.366	0.503	0.53	0.839	1	0.974	0.95	0.925	0.728	0.575	0.448	0.723	0.801
D3(520-600)	0.639	0.673	0.692	0.441	0.537	0.525	0.814	0.974	1	0.985	0.975	0.807	0.676	0.569	0.77	0.793
D4(605-625)	0.665	0.7	0.728	0.46	0.559	0.538	0.818	0.95	0.985	1	0.978	0.796	0.669	0.593	0.799	0.802
D5(630-690)	0.585	0.636	0.688	0.497	0.581	0.529	0.75	0.925	0.975	0.978	1	0.836	0.728	0.643	0.83	0.806
D6(695-750)	0.453	0.526	0.551	0.664	0.445	0.308	0.621	0.728	0.807	0.796	0.836	1	0.964	0.794	0.644	0.532
D7(760-900)	0.368	0.451	0.479	0.71	0.429	0.25	0.496	0.575	0.676	0.669	0.728	0.964	1	0.897	0.604	0.413
D8(910-1050)	0.397	0.475	0.508	0.682	0.498	0.319	0.422	0.448	0.569	0.593	0.643	0.794	0.897	1	0.656	0.372
D9(1550-1750)	0.483	0.526	0.594	0.448	0.693	0.611	0.582	0.723	0.77	0.799	0.83	0.644	0.604	0.656	1	0.885
D10(2080-2350)	0.461	0.469	0.519	0.247	0.582	0.628	0.633	0.801	0.793	0.802	0.806	0.532	0.413	0.372	0.885	1

Table 2: Correlation matrix for Landsat TM acquired on 23.08.94 and Daedalus acquired on 24.08.94. Red denotes bands that do not overlap with Landsat TM.

Figure 15: LEFT: Raw bands 4 (605-625nm), 6 (695-750nm), 8 (910-1050nm). RIGHT: Corrected bands 4 (605-625nm), 6 (695-750nm), 8 (910-1050nm).

The following DMSV images were acquired over Opthalmia Ranges in the Pilbara, WA, and were used to test the referencing method for DMSV data with bands that do not overlap those of Landsat TM. There is a time difference of 17 months between the two acquisitions.

Figure 16

shows bands 1, 2, 3 and 4 of a DMSV image acquired on 01.10.95. For each of the DMSV bands, the Landsat TM band that is closest to that DMSV band was used to model the brightness surfaces on Figure 17. Figure 18 shows the brightness surfaces for the DMSV bands modelled using a combination of all the Landsat TM bands. Figure 19 show bands 1, 2, 3 and 4 of the adjacent DMSV image. For each of the DMSV bands, the Landsat TM band that is closest to that DMSV band was used to model the brightness surfaces on Figure 20. Figure 21 shows the brightness surfaces produced using a combination of all the Landsat TM bands to model the DMSV bands.

Table 3: Correlation matrix for Landsat TM acquired on 24.03.94 and DMSV acquired on 01.10.95. Red denotes bands that do not overlap with Landsat TM.

Figure 16: Raw bands 1, 2, 3 and 4 for DMSV image o5ac.

Figure 17: Brightness surface for bands 1, 2, 3 and 4 for DMSV image o5ac modelled using the closest Landsat TM band.

Figure 18: Brightness surface for bands 1, 2, 3 and 4 for DMSV image o5ac modelled using all the Landsat TM bands.

Figure 19: Bands 1, 2, 3 and 4 for DMSV image o5ad.

Figure 20: Brightness surface for bands 1, 2, 3 and 4 for DMSV image o5ad modelled using the closest Landsat TM band.

Figure 21: Brightness surface for bands 1, 2, 3 and 4 for DMSV image o5ad modelled using all the Landsat TM bands.

Figure 22: These images were produced using band 4 in red, band 3 in green and band 2 in blue. LEFT: Mosaic of 2 uncorrected images, o5ac on o5ad. MIDDLE: Mosaic of 2 images corrected using the closest Landsat TM as reference, o5ac on o5ad. RIGHT: Mosaic of 2 images corrected using all the Landsat TM as reference, o5ac on o5ad.

Figure 23: These images were produced using band 3 in red, band 2 in green and band 1 in blue. LEFT: Mosaic of 2 uncorrected images, o5ad on o5ac. CENTRE: Mosaic of 2 images corrected using the closest Landsat TM as reference, o5ad on o5ac. RIGHT: Mosaic of 2 images corrected using all the Landsat TM as reference, o5ac on o5ad.