A Knowledge Based Classification for Urban Mapping Using High Resolution Remote Sensing Data

This paper addresses the weakness of pixel based or cell-based classification algorithms for urban mapping. They cannot provide a handy object level classification results as often preferred in urban planning and assessment applications. A knowledge based approach is proposed which is integrated with the spectral classification methods and image processing tools. Experiments using a high resolution image were conducted. The results show the proposed method can improve the performances obtained by using the conventional maximum likelihood or ECHO algorithms. Key words: Knowledge based, Object attributes, classification.