Automatic Interpretation of Vegetation Areas in Brazil

This paper describes a procedure for a multitemporal interpretation of vegetation areas in Brazil using Landsat TM images. We started with a monotemporal interpretation of the first epoch of the scene by using both spectral and structural features. For interpretation of temporal changes we discretely described temporal conditions of regions, and transferred the most probable temporal changes of the given conditions as temporal knowledge into a state transition diagram, which was used for the multitemporal interpretation of the images. This diagram contains limitations of possible state transitions derived from biological reasons as well as limitations given by law. We included the interpretation strategy into the image analysis system eCognition. In this paper we describe how we created the state transition diagram for our test area and how we included it in eCognition. By using temporal knowledge it is possible to reduce the search space during the multitemporal interpretation and reduce therefore the classification errors. The results show that the presented procedure is suitable for multitemporal interpretation of that area, and leads to better results than the procedures used for those areas so far.