| Popis: |
Satellite imaging has numerous applications in several fields such as urban planning, disaster detection and recovery, agriculture development, 3D GIS, and many more. One of the major challenges of satellite imaging is to extract the building footprints that are vital for disaster planning and recovery. Existing algorithms retain quantitative scores; however, they fail to preserve shape of building footprint. To address this problem, we propose here a building footprint extraction algorithm from satellite images; our objective is to formulate the building pixels in an efficient way, exploiting eccentricity parameter, in order to extract the exact shape of buildings. The proposed hybrid approach comprises of superpixel generation and parameterization such as axis ratio, eccentricity, color, and segment area which have been used to extract the true building footprint. The qualitative and quantitative evaluations demonstrate that the proposed algorithm can extract the building footprint more efficiently over other existing methods, in terms of correctness, completeness, and F-measure. In addition, we also evaluate the Dice similarity coefficients of extracted building footprints, which shows the influence of color parameter for effective extraction providing better results. |
