Here we present an innovative method to automatically detect a valuable set of landscape patterns within a satellite image. We are aiming to achieve two major goals. The first deals with capabilities to capture the knowledge shared by experts for image analysis tasks. The second emphasizes ontology supporting roles to retain and reuse knowledge in an extendable and interoperable way. The paper highlights the synergistic use of different strategies including image preprocessing, operations using specific metrics, inductive logic programming (ILP) and knowledge modelling leading to a more global approach. In particular we explore the interplay between an inductive (ILP) and a deductive method (ontology engineering) to fully meet these objectives.