Classification of croplands through integration of remote sensing, GIS, and historical database

Abstract

This work presents a methodology to classify croplands using a multitemporal/historical dataset of images and ground ancillary data referring to three consecutive years. An image processing/geographic information system as well as a database management system (DBMS) were used to make the integration of these multisource data. In order to evaluate the usefulness of a database for crop classification, the area under study was digitally classified by two groups of interpreters, using two methodologies: (a) the proposed methodology using maximum likelihood classification assisted by an historical/multisource database, and ( b) a conventional maximum likelihood classification only. Both results were compared using the Kappa statistics. The indices to both the proposed and the conventional digital classification methodologies were 0.669 (very good) and 0-472 (good), respectively. The use of the database rendered an improvement over the conventional digital classification. Furthermore, along with this study some problems related to multisource data integration are discussed.