Figure 4. Landsat TM data for selected site, bands 7, 5, 4 in colour 
bands blue, green and red respectively. 
Figure 5. Class label image derived from analysis and 
classification of Landsat TM data depicted in figure 4. Of interest 
is the hardwood class depicted in green. 
Figure 6 Hardwood plantation boundary from inventory database. 
5 Results / conclusion 
The model was applied to the data producing the results depicted 
in figures 7 and 8, which may be compared with the pre model 
values shown in figures 5 and 6. 
From figure 5 and 7 we observe that the class label image has 
more uniform regions after the application of the model, though 
the relative contribution of the neighbourhood and boundary 
priors was not evaluated here. 
From figures 6 and 8, we observe that the hardwood boundary 
estimate is improved, though still in error. 
The difficulty of updating a boundary in a relatively cluttered DT 
surface (figure 9) requires further attention, along with the 
specification of the relative weights of each component of the 
model. 
Figure 7. Resulting class label image produced using the model. 
/L 
Figure 8. Updated boundary estimate. Although a better estimate, 
a portion of the left side of the boundary has not been well 
corrected. 
Figure 9. Distance transform of class label boundaries for final 
iteration. This defines the topographic surface used for boundary 
movement. 
ACKNOWLEDGEMENTS 
This work has been supported by the Australian Government 
Department of Environment and Heritage, through management 
and coordination of the National Carbon Accounting System 
Landcover Change Project, details of which can be found at 
http://www.greenhouse.gov.au/ncas/.