The influence of urban structures on impervious surface maps from airborne hyperspectral data.

摘要: Abstract Airborne hyperspectral data fulfills the high spectral and spatial resolution requirements of urban remote sensing applications. Its information content enables delineating impervious areas including separation built-up non surfaces, thus being relevance for many environmental However, two phenomena related to surface structure negatively impact accuracy maps from such airborne sets: (1) displaced buildings that lead confusion between class adjacent as a function building height view-angle; (2) street trees obscuring underneath. Both effects have so far not been investigated potential sources inaccuracy are usually differentiated in analysis utilizing data. Thus, positive influence might undervalued cases. We set up an scheme allows separately quantifying error when producing land cover areas. Given reliable cadastral on extent network, detailed relatively large Hyperspectral Mapper acquired over Berlin, Germany, was performed. Results show both displacement obscured by tree crowns great impact: at view-angles, adds 16% compared nadir regions; more than 30% area is classified vegetation. Moreover, irregularities prohibit empirical correction: misclassification due also depends view-direction, i.e. illumination properties shadow, while differs significantly along streets inside residential this work underline necessity consider all image processing steps evaluating reliability products they depict directions future methodological development.