RADARSAT-2 fine-beam polarimetric and ultra-fine beam SAR data for urban land cover mapping: Comparison and Synergy
(English)Manuscript (preprint) (Other academic)
This paper investigates the capabilities of multitemporal RADARSAT-2 fine-beam polarimetric SAR (PolSAR) data and ultra-fine beam C-HH SAR data for the detailed urban land cover mapping using a novel contextual approach. With an adaptive Markov Random Field (MRF) and the spatially variant Finite Mixture Model (FMM), contextual information was effectively explored to improve the mapping accuracy. The results showed that the contextual approach could produce homogenous classification while preserve shape details. Compared with C-HH SAR, PolSAR data were important for identify various urban patterns. Nevertheless, efficiency of the C-HH SAR textures for extraction of the built-up area was observed. Thus we proposed a texture enhancement in FMM to further improve the classification accuracy. Moreover, a rule-based approach employing object features and spatial relationships has been used to extract the road, street and park with reasonable accuracy. Three-date RADARSAT-2 fine-beam PolSAR and three-date ultra-fine beam C-HH SAR data over the Greater Toronto Area were used for the evaluation.
Adaptive MRF, polarimetric SAR, texture, urban land cover.
IdentifiersURN: urn:nbn:se:kth:diva-104760OAI: oai:DiVA.org:kth-104760DiVA: diva2:567139
QS 20122012-11-122012-11-122012-11-12Bibliographically approved