Hyperspectral imagery and urban green observation

Urban vegetation is known to play a major role in improving urban environmental conditions and to contribute to urban life quality. Consequently, nowadays most central European cities take special care on green spaces management and research on the effects is performed in different disciplines. Both need precise and up-to-date information, also beyond public areas. Using remote sensing is therefore valuable and especially since high resolution data are available for urban studies. Hyperspectral data are potentially interesting for such highly heterogeneous areas. However, there has been limited evaluation of the use of hyperspectral images and processing techniques for mapping urban vegetation. Here, we discuss the potential to use hyperspectral data for vegetation detection in urban areas. For our study we analyze field spectra and an image acquired by the Compact Airborne Spectral Imager (CASI) in September 2005 over Strasbourg city area (France). We present the first results on evaluation of coherence between field and image data. For image processing, a vegetation mask and minimum noise fraction transformation (MNF) is applied to reduce and compress the data. The following Spectral Angle Mapper (SAM) classification aims at vegetation identification at species level. We finally discuss our experiences and illustrate the difficulties encountered with urban vegetation detection.