A linearised pixel-swapping method for mapping rural linear land cover features from fine spatial resolution remotely sensed imagery

摘要: Accurate maps of rural linear land cover features, such as paths and hedgerows, would be useful to ecologists, conservation managers planning agencies. Such information might used in a variety applications (e.g., ecological, management applications). Based on the phenomenon spatial dependence, sub-pixel mapping techniques can increase resolution produced from satellite sensor imagery map features with increased accuracy. Aerial photography 0.25m was acquired Christchurch area Dorset, UK. The hard classified using simple Mahalanobis distance classifier classification degraded simulate proportion images resolutions 2.5 5m. A pixel-swapping algorithm then applied each images. Sub-pixels within pixels were swapped iteratively until correlation between neighbouring sub-pixels for entire image maximised. Visual inspection super-resolved output showed that prediction position dimensions hedgerows comparable original imagery. displayed an accuracy 87%. To enhance output, anisotropic modelling component added. direction largest sums proportions calculated moving window at pixel level. orthogonal sum estimating anisotropy ratio. ratio modify so likelihood creating map. new method led fine approximately 5% compared conventional method.