Abstract
This paper deals with application of a new model established by the Kalman filtering theory to the land-cover classification of Landsat TM data. The results compared to existing three classification techniques showed that classification accuracy could be improved in average by using the Kalman filtering model. Furthermore, this model has a considerable advantage in real world application, being not limited by the spatial resolution of remotely sensed data and making it possible to carry out classification when there are many land-cover categories contained in a small area unit having the size of the order of 200m×200m.
