The geometric correction was performed to the Landsat 7 August images because a great number of details, due to less vegetation cover, were more easily recognised. The same set of Ground Control Points (GCPs) were subsequently relocated on the other images. The four scenes were consecutively cropped according to the study area co-ordinates. The EASI/PACE v7.0 software was utilised for image processing.
The next step of the image processing was focused on the analysis of the most suitable False Colour Composite (FCC) image for visual interpretation. For vegetation and soil study it was preferred to discard the true colour image 321 (RGB) due to low information given.
The following seven colour composites (RGB) were compared and analysed:
Independently from the month, in 432 FCC (the image is like an infrared film camera) vegetation appears with different hue of red and bare soil appears with a variety of colours according to soil characteristics and rock types. In Figure 5 the soil types can be easily distinguished; in the upper part of the image the red pisolitic silts dunes appear in green colour and the hydromorphic soils are whitish.
The 543 FCC is useful for forest study: vegetation appears green and stressed vegetation appears brownish, facilitating differentiation. The FCC looses information from soil types which are less separable because the image looses information from Band 2 where soils have a high reflectance (Figure 6).
|
Figure 5 - 432 RGB of 08/2000 |
Figure 6 - 543 of 08/2000 |
The 457 FCC could be utilised to stress the difference in soil moisture due to Band 7 and Band 5 where, according to the increase of water content, the reflectance decreases emphasising peculiar characteristics of vegetation (Figure 7). The B4 maintains information coming from vegetation that appears in red-orange colour where vegetation is healthy and in brown where vegetation is stressed.
The 54NDVI FCC was an attempt to show vegetation changing throughout the year using the information given by the NDVI channel. Good results were achieved for vegetation study but the obtained image presented too many colours that weren’t useful for soil identification and for the field maps (Figure 8).
|
Figure 7 - 457 of 08/2000 |
Figure 8 - 5 (08/2000), 4 (08/2000), NDVI (04/2000) |
The 432 FCC on August was chosen as the image for photo-interpretation and for the field work map on the scale of 1:50,000 because vegetation cover and soil type were more detectable.
To increase the ground resolution from 30m to 15m, fusion methodology was applied. Fusion performs data fusion of a true colour or pseudocolour image with a black and white image; different procedures can be applied to obtain the fusion and the one adopted was the Brovey transformation (Ranchin and Wald, 2000). This one (Table 8) divides the Bands to display in a given colour layer (in this case Band 4, Band 3, and Band 2) and then multiplies by the intensity layer (Panchromatic channel) The Brovey algorithm is a highly effective transformation that generates a better looking image if compared to the direct substitution of I (Intensity) channel with the Panchromatic one. To Band 3 and Band 2 of the fused image a 3*3 edge sharpening filter was then applied to decrease noises; to B4 no filter was applied because it was not effective in noise reduction. Image prints on the scale of 1:50,000 were produced for field use.
|
R= B4; G= B3; B= B2; I= Panchromatic band |
||
|
Red layer |
Green layer |
Blue layer |
|
[R/(R+G+B)]*I |
[G/(R+G+B)]*I |
[B/(R+G+B)]*I |
Table 8 - Brovey transformation formula for the fusion procedure