Spectral reflectance data are central to the mapping of urban surfaces and materials from satellite and aerial imagery. The number of available samples is very small for the variability of materials within each category (e.g., asphalt). Many factors impact surface reflectance, because of material composition linked to national/local regulations related to weather/climate conditions, local availability of raw materials and socio-economic factors. To help address this data-gap, the urbisphere urban Hyperspectral library is being developed, using RS-3500 spectroradiometer. The RS-3500 spectroradiometer captures spectral radiation from 350nm to 2500nm with 1nm spectral resolution. This will use observations taken in several cities.

The urbisphere hyperspectral measurements are organized to clusters (groups of HS measurements with similar reflectance spectrum). Each measurement is delivered with its own image, measurement ID, spectral reflectance, unique material name/color/structure/status/usage/location, spectral incoming radiation, spectral outgoing radiation and metadata for the quality and the conditions during the measurements. The spectra are also adjusted to several Satellite sensors (Sentinel-2, Landsat-8, Planet SuperDove, WorldView-3). The adjusted spectra can be used directly for ML training for the respective L2A satellite images. The first samples are available below.

xlsx mat

xlsx mat

xlsx mat