Form pottorns 
In image data of high spatial resolution with regard to the 
evaluation unit, there are many classes {ez forestry, housing 
areas) which are represented by single objects (trees, buildings) 
of sufficient number and typical shape. To interpret this data 
it can be of advantage to predetermine the patterns of those 
single objects. After having classified these objects, there 
are decision rules to characterize the evaluation units based 
on the presence or absence of typical objects. Patterns to 
classify the objects are for example the surface, the perimeter, 
or the axis of inertia. Often the ratios between these features, 
like the surface versus the perimeter or the surface versus the 
main axis of inertia are powerful form patterns, too. 
To gain these patterns a preprocessing procedure is needed which 
extracts the single objects out of the grey level data. This is 
normally done by image binarisation algorithms like level slicing 
or thresholding of constant, locally adaptive or object dependent 
type.