KNOWLEDGE-BASED TECHNIQUES FOR MULTISOURCE CLASSIFICATION

The value of utilizing multiple data sources for classifying images has long been recognized in remote sensing. However, any attempts to do so have faced enormous problems primarily due to the inadequacy of traditional single source analytical techniques. This paper demonstrates the feasability of using knowledge-based procedures to provide a new scheme for incorporating several sources in the classification process. The two schemes presented (based on numerical and qualitative reasoning) are computationally efficient and have high classification accuracies. © 1990, Taylor & Francis LLC.