Identification of Urban Features Using Object-oriented Image Analysis

The objective of this work was the presentation of a knowledge-based framework for the identification of urban features, as well as additional land cover types, contained in the image scene. The test area used was located in an urban environment of Aghios Stefanos, Attica. An IKONOS image (spatial resolution of 1m-resampled) was acquired for the study area and further pre-processed. Furthermore, supplementary geo-data were created and embedded in the knowledge-based system focused on the identification of urban characteristics (e.g. building boundaries), such as derived thematic data from remote sensing techniques and the constructed DSM of the study area (with a spacing of 3m), which was extracted using photogrammetric methods, for providing the required 3-D information. The knowledge-based methodological framework consisted of: (1) Problem conceptualization and knowledge acquisition, which involved the identification of urban features and surrounding land cover classes faced by “decoding” of knowledge based on the existing data types and context. Furthermore, remote sensing analysis methods, implementation of edge detection algorithms and multi-scale hierarchical segmentation were investigated and applied. (2) Knowledge representation, which was performed using object-oriented image classification using fuzzy representation models (proper Fuzzy Membership Functions (FMF) or Nearest Neighbor Classification (NNC)) for the assignment of objects into the defined classes. The output result of the proposed knowledge-based identification system is expected to be a final map that contained urban features of interest (such as building boundaries and road segments) and additional land-cover classes which consist the natural scene of the study area (e.g. vegetated regions).