DOI:10.20894/IJDMTA.
Periodicity: Bi Annual.
Impact Factor:
SJIF:4.893 & GIF:0.787
Submission:Any Time
Publisher: IIR Groups
Language: English
Review Process:
Double Blinded
Extracting Building features from satellite imagery is a vital research area in digital remote sensing field. Building detection & extraction is one of the complex and challenging task in GIS database. But this technique is useful for urban planning and obtaining more timely and accurate information during natural disasters like Earth quakes, Cyclone etc..At the first appearance, buildings are visible as easiest objects to detect and extract. But many difficulties meet on extracting buildings accurately, comprising various outlook angles, roof top complexity, environmental objects(Trees, Roads, vehicles etc) and additional objects which ambiguous the boundaries of the buildings that can be detected. Because of these reasons many algorithms deliver less quality of building extraction and also much time taken for detection. To solve this problem, integrating many efficient algorithms provides better results than individual algorithms. In this study, an enhanced approach for rising the quality and accuracy in detecting & extracting building textures with various and complex angles of roofs from urban area satellite images is proposed. First, an unsupervised image segmentation approach based on SOM(Self organizing maps) is applied to detect roof top regions. Then, the SOM combines with MRF(Markov Random field) spatial constraints for improved segmentation outcomes. This Hybrid approach of SOM and MRF used less data samples in training set. Experimental results obtained that the proposed method achieved excellent result in detecting and extracting rooftops in complex satellite images.
