onal 
ns 
5% 
er 
ds 
ation 
= 329 - 
example we have chosen we had to classify only 64 x 64 possible vectors, 
filling with zeros the rest of the table. The result of such classification 
is the table shown in Figure l. This table can be displayed on a CRT screen 
and thus the user can appraise the result of the training process which 
determines the division of the feature space into several classes. For 
example in Figure 1 we see that the classes B and S are disputing the 
same area and conclude that it is not possible to differentiate between 
the corresponding types of ground cover and take the decision to define 
a new class which is the union of them. 
It is clear that the idea to classify all possible vectors to generate 
the table can be achieved with any other decision rule. In future work we 
intend to consider possible alternatives to the maximum likelihood rule, 
with the main objective of dropping the assumption that the classes are 
normally distributed. As a help for this research project we have introduced 
another feature in the system which consists in the possibility of modifying 
the table interactively, asigning a value to a given member of the matrix 
or to several neighbouring members. 
4. CONCLUSIONS 
In a situation where the user wants to classify a large area into an 
small number of general classes, the use of 2-dimensional tables is very 
helpful, saving a lot of time. This is made possible by the circumstance 
that LANDSAT data is essentially 2-dimensional. The display and interactive 
modification of classification tables may help the users's understanding 
of the classification process and may help the investigation of classification 
criteria. 
5. BIBLIOGRAPHY 
[11 S. Shlien and A. Smith, A rapid method to generate spectral theme 
classification of LANDSAT imagery, J. Rem. Sensing Envir. 4, 67-77 
(1975). 
[27 P.N.Misra and S.G. Wheeler, Crop Classification with LANDSAT multi- 
spectral scanner data, Pattern Recognition, 10, 1-13 (1978). 
[3] M. Rebollo, F. Ortí and J.M.Camarasa, Supervised and Unsupervised 
Classification of the Delta of the Ebro River: Land Use Study using 
LANDSAT Data, Centro de Investigación UAM-IBM, SCR-01.77 (1977)