Research Methodology - Historical Land Use/Land Cover Classification Using Remote Sensing - page 145

Fig. 5.59 Illustration of the individual 16-project-based classification results (permanent and
winter crops classification results for 1987)
set was then mosaiced to produce only two mosaic-images, each representing one
date. Again, iMAD was applied to the two resulting mosaic-images. The master-
scene was chosen as the mosaic-scene produced from the TM-data in 2007, as it
was possible to get weather data for the dates, and radiometric normalizations and
atmospheric corrections were able to be performed. Finally, the two mosaic-scenes
were geometrically registered using the image to image method. The MSS-scene
was also resampled to the same spatial resolution of 30 m as TM-scene. Finally,
the two remote sensing data-scenes were added to the change detection mapping
process using the image differencing method. The three major mapped changes
over the last 32 years were: natural areas to bare areas, bare areas to cultivated
areas, and no change (see Chap. 6.3.1 ).
5.12.2 Post-Classification Approach
The post-classification change detection approach concerned the analysis of the
differences between two more or less independently classified images. A com-
parison of the categorizations was performed using raster-based analysis (ENVI, v.
4.6). The major merit of this approach was that data normalization was not needed
because the remote sensing data recorded at two dates were classified separately