Combining Object-Based Classification and Data Mining Algorithm to Classify Urban Surface Materials from WorldView-2 Satellite Image

Although object-based image analysis (OBIA) has been used for detailed classification of urban areas, its attribute selection and knowledge discovery have been time consuming and subjective to analysts´ performance. In this study, Data Mining was performed using C4.5 algorithm to select the appropriate attributes for object-based classification. This algorithm provides a decision tree output to represent the knowledge model, enabled a faster classification of intra-urban classes, and disabled the subjectivities which are related to the interaction of the analyst. The decision tree results were implemented in eCognition software to provide an effective and fast generation of semantic network in OBIA.