Abstract
This paper develops a method for analyzing trend curves that is robust against errors and noises. First, we classify the local. topological characteristics of a trend curve in three categories: a "peak", a "bottom" and a "slope". Second, we define "distinctiveness" of these topological categories. Third, we describe the characteristics of a trend curve using a configuration of these peaks and bottoms with the distinctiveness, and introduce the concept of "isomorphic" relations with respect to this configuration. We develop a method of classifying trend curves and define the similarity between two trend curves using the isomorphic relations. Finally, we apply the proposed method to the analysis of land cover change in the Persian Gulf region using trend curves of NDVI obtained from NOAA remotely sensed data over a year from 1982-1993.
