Land Cover Mapping In Urban Environments Using Hyperspectral Data

With the rapid development of urbanization and the continuous improvement of people's living standards,more people live in cities.For dynamic administration,fast and accurate urban land cover mapping is needed.Land cover classification refers to labeling each pixel in the image through a supervised or unsupervised classification algorithm,deriving urban land cover maps and eventually extracting useful land cover information based on user needs.Thanks to the successful launch of a series of satellites,remote sensed images collected by the sensors carried by satellites have become the main data source for extracting land cover information.With the increasement of the spectral resolution of sensors,hyperspectral image classification has attracted experts and scholars' attention in the field of remote sensing.But the widely distributed sparsely vegetated surfaces in urban areas generally cause difficulties to map soil and other soil-related land cover.In addition,the shadowing problem that relates to urban surface structure negatively impacts the accuracy of urban land cover maps.Both effects have so far not been sufficiently investigated for airborne hyperspectral data.In this paper,a new classification framework for mapping land cover in urban environments using high spatial resolution hyperspectral data was proposed.The proposed classification scheme was applied to map urban land cover using APEX data in the city of Baden,Switzerland.Firstly the NDWI and NDVI indices are used to separate the land cover in the scene into three main classes:water,vegetation and non-vegetated surface.Then the vegetation and non-vegetated surfaces are separately classified into more detailed ground categories based on object-oriented methods.It is fuzzy to label the sparsely vegetated soil as soil or vegetation because its vegetation abundance has a continuous gradation.Soil was classified both in vegetation and non-vegetated surface,and these two soil results were merged in the final classification map.Shadows usually occur in urban landscape,which results in the underestimation of shaded ground materials.Shadows were initially classified in shaded vegetated surfaces and shaded non-vegetated surfaces,and then they were further classified into meaningful ground categories.To evaluate the performance of the framework,several comparative experiments were carried out.The experimental results demonstrate that the proposed classification framework is well suited for mapping land cover in urban environments using high resolution hyperspectral data.Results from this work underline the necessity to consider the accuracy of soil when evaluating urban environments using remote sensing data.Results from this study also demonstrate that although shaded surfaces are generally classified as a single category in urban environments,in high resolution hyperspectral data,the shadows can be further classified into meaningful land cover classes with an acceptable accuracy.