Feature recognition techniques

Abstract

Almost all applications of remotely sensed imagery require generic algorithms for image feature extraction and classification to gain the required information. Therefore the GMOSS project defined a work package Feature recognition to serve the application work packages in their need to derive information for their tasks. For this purpose an important task is the definition of terms, nomenclature and the creation of a feature catalogue which describes significant features as well as the ability and means to detect these features. The work performed in this work package covers a very wide area and reaches from basic image processing algorithms used in pre-processing steps to highly sophisticated automated, object-based classification and detection methods and its evaluation regarding to performance and robustness. In principle two basic operations will be covered by the feature recognition work package. Classification should provide good and robust background knowledge of the basic land-cover within a certain area whereas object detection techniques are specialized on finding one specific feature or object in a defined area. © 2009 Springer Science+Business Media B.V.