This model was designed to produce a layer-stacked composite image to be used for segmentation purposes, the particular variables involved having been determined to be optimal for producing segments appropriate to vegetation mapping. Increased segmentation accuracy in turn bolsters the accuracy of any subsequent OBIA. The model calculates NDVI from NAIP imagery (assuming Red is band 1 and NIR is band 4), Trishade (aka fully illuminated hillshade) from a DEM, the first Principal Component (reflectivity of surfaces from all band inputs) from a Landsat image, and stacks these calculated layers with the original 4 bands of the NAIP image. A Spectral Attribute File (.saf) is also provided to make it easier to select predefined band combinations for displaying the stacked data. Drop the .saf into your $IMAGINE_HOME\etc directory (and restart ERDAS IMAGINE) to access these band combinations.
Note that all input imagery has been resampled to a common pixel size (usually 10m). Slight modifications may be required if the inputs vary in pixel size, such as selecting more appropriate resampling techniques.Note that the original design makes several assumptions:
ForestryLayerstack-v15-1-0.gmdx |
Stretched NDVI submodel |
Modified Principal Components submodel |
Trishade submodel |
NAIP Filename In: NAIP input image; assumes band order is RGBN.
Landsat Filename In: Landsat input image for producing modified principal components.
Stats Skip Factor: Larger numbers will increase processing speed at the expense of some accuracy. 1 uses all pixels values, 4 uses every fourth row and column value, etc.
DEM Filename In: DEM input for producing trishade layers.
Layerstack Filename Out: Name of layerstack output image.
Shown below is the trishade portion of the 9-band output image, RGB set to bands 6,7,9: |