1. Use DISPLAY to view these images. Using autoscaling (by clicking on the autoscaling control box) and the GREY256 palette will achieve a simple linear contrast stretch and grey level image.
2. Use STRETCH with the seven bands, select linear with saturation option, specify 5% saturation and accept the defaults. Display the results using autoscaling and the GREY256 palette.
3. To choose training sites, let's create a false color composite image first. Run COMPOSIT from the Analysis/Image Processing menu to create a false color composite image using band 3 as the blue, band 4 as the green, and band 7 as the red bands. Specify linear stretch with 0.5% saturation and 256 levels. Do not exclude 0 from the stretch. Use DISPLAY with Composit256 palette to view this image. Vegetation should appear green, with cropland light green and forest very dark green.If you are using the smaller subset images, there will be just small patches of forest present. If you are using the full data set, you will notice major stands of forest in the northwest. Water will apppear blue or black. Urban landscape will appear a mottled mix of pink and purple. Barren fields will appear light pink and white.
Create training site files for the five land-cover types using on-screen digitizing option. Once you have vector files ready, run MAKESIG to create a signature file for each land cover. Next, run MAXLIKE to perform supervised classification. Specify the five signature files, 5% of pixels to be excluded from classification, and select equal prior probabilities for each signature. Use all seven bands in the classification. Display the result with QUAL16 palette. Compare this image with the false color composite.
Run AREA on the classified image, and select tabular output in pixels. Record the number of pixels assigned to each class.
4. Use DISPLAY to view the classified image with QUAL16 palette.
Return to the Digital Analysis of Northern Italy