Direkt zum InhaltDirekt zur SucheDirekt zur Navigation
▼ Zielgruppen ▼

Faculty of Mathematics and Natural Sciences - Geomatics

Berlin II - Urban environmental monitoring with high resolution remote sensing data

(Städtisches Umweltmonitoring mit spektral und geometrisch hoch auflösenden Fernerkundungsdaten)

Project period: 03/2009 - 12/2012

Descrition of the project

The scientific focus of this project was to characterize the city of Berlin with imaging spectrometry data and auxiliary information such as height information. In the first part of the project, several case studies were carried out where different image data setups were analyzed with different advanced processing techniques. A special emphasis was put on the potential and limitations for accurate per-pixel-based mappings of various man-made materials as well as of urban vegetation types and their physiological conditions. The second part of the project was closely linked to the activities of the EnMap Core Science Team with regard to ecosystem transition. The goal was to develop a universal modeling approach for a sub-pixel based quantification of ecologically meaningful land cover types along the urban-rural gradient. To investigate the methods’ suitability for imagery from future hyperspectral satellite missions, airborne and simulated spaceborne imaging spectrometer data were analyzed at different spatial resolutions.

Results

The case studies of this project were successfully published in international peer-reviewed journals/books and contributed to several international conferences (publications). In summary, the following major conclusions can be drawn from the achieved results:

  • The high spectral information content of imaging spectrometer data can be well exploited by machine learning algorithms for a precise per-pixel classification of urban areas. In most cases an accurate identification of typical man-made materials and vegetation types was possible.
  • The integration of height information and height metrics into the processing scheme helped to overcome limitations of analyses based on a purely spectral basis. Requirements for the use of different datasets for qualitative mapping assessments were identified and recommendations for future studies were given.
  • Accurate urban land cover fraction maps were derived through the combination of imaging spectrometer data, synthetically mixed training information from a spectral library and kernel-based regression modeling techniques. The developed approach can be flexibly adapted to the spectral diversity and the mix pixel problem typical for urban remote sensing data. Thus, a first step towards the realization of a universal quantitative model for urban spaces was taken.
  • Results achieved when transferring the models from airborne to spaceborne scales indicate high potential of future hyperspectral satellite mission EnMAP for urban land cover assessments.

Simulated EnMap Satellite images - Berlin training data

Impervious (red), vegetation (green) and soil (blue) fraction maps derived from (a) simulated EnMAP, and (b) HyMap and Landsat data using the combination of Support Vector Regression and synthetically mixed training data. Source: Okujeni et al. (2015)

Publications

Peer-reviewed papers
van der Linden, S., & Hostert, P. (2009). The influence or urban surface structures on the accuracy of impervious area maps from airborne hyperspectral data. Remote Sensing of Environment, 113, 2298-2305.

Hostert, P. (2010). Processing Techniques for Hyperspectral Data. In Rashed, T. & Jürgens, C. (Eds.), Remote Sensing of Urban and Suburban Areas (pp. 165-179). Springer.


Okujeni, A., van der Linden, S., Tits, L., Somers, B., & Hostert, P. (2013). Support vector regression and synthetically mixed training data for quantifying urban land cover. Remote Sensing of Environment, 137, 184-197:

Okujeni, A., van der Linden, S., Jakimow, B., Rabe, A., Verrelst, J., & Hostert, P. (2014). A comparison of advanced regression algorithms for quantifying urban land cover. Remote Sensing, 6, 6324-6346.

Okujeni, A., van der Linden, S., & Hostert, P. (2015). Extending the vegetation-impervious-soil model using simulated EnMAP data and machine learning. Remote Sensing of Environment, 158, 69-80.

Non Peer-reviewed conference contributions

Okujeni, A., van der Linden, S., Cierpinski, M., & Hostert, P. (2013). Transferring Support Vector Regression Models from airborne to EnMAP-like data from urban areas. 8th EARSeL Workshop of Special Interest Group in Imaging Spectroscopy. Nantes, France, 08-10 April 2013.

Okujeni, A., van der Linden, S., Cierpinski, M., & Hostert, P. (2013). A comparison of advanced regression algorithms for quantitative mapping of urban land-cover. Dreiländertagung D - A - CH 33, Wissenschaftlich-Technische Jahrestagung der DGPF. Freiburg, Germany, 27 February - 01 March 2013.

Okujeni, A., van der Linden, S., Rabe, A., Cierpinski, M., Hostert, P. (2012). On the influence of spatial scale on the accuracy of support vector regression models from synthetically mixed spectral libraries. IEEE Geoscience and Remote Sensing Symposium (IGARSS). Munich, Germany, 22-27 July 2012.

Okujeni, A., van der Linden, S., Cierpinski, M., Hostert, P. (2011). A non-linear approach to describe the spectral mixing space of urban areas. 7th EARSeL Workshop of Special Interest Group in Imaging Spectroscopy. Edinburgh, Scotland, UK, 11-13 April 2011.

Rabe, A., Jakimow, B., van der Linden, S., Okujeni, A., Suess, S., Leitão, P., & Hostert, P. (2013). Simplifying Support Vector Regression Parameterisation by Heuristic Search for Optimal e-Loss. In, 8th EARSeL Workshop of Special Interest Group in Imaging Spectroscopy. Nantes, France , 08-10 April 2013.


Lab members involved
Funding

This project was funded by the German Research Foundation (DFG)