(Den Text gibt es leider nur auf Englisch)
Monitoring of the urban environment as well as modeling of biotope types requires detailed knowledge about the surface materials. Due to the problem of a high percentage of spectrally mixed pixels, there is a great need for new algorithms for automatic image analysis which are capable of dealing with the special conditions in urban areas. A suitable processing chain is under development that has already reached a high degree of automation. It consists of a material identification and classification process followed by an iterative linear spectral unmixing procedure resulting in surface fraction layers for each material.
Current research is focused on the optimization of the automated material identification process. The presently used classifier is capable to process reflectance values as well as numerical features derived from the spectra contained in the spectral library. Optimal input features are automatically identified by the one-against-one classifier during the training phase optimizing the separation of two classes. The additional use of numerical features improves the robustness of the algorithm against yet unknown spectral variability in new image data. Identified materials are subsequently used to identify spectrally pure pixels and representative spectral endmembers. Based on these information an iterative neighborhood-oriented linear spectral unmixing procedure is used to calculate the surface fraction layers. It benefits from the knowledge of neighboring pixels identified as spectrally pure ones in the process of selecting meaningful endmember combinations possessing a high likelihood for the representation of the actual composition of mixed pixels.
Heiden, U.; Segl, K.; Roessner, S.; Kaufmann, H.: Determination of robust spectral features for identification of urban surface materials in hyperspectral remote sensing data. In: Remote Sensing of Environment, 111, 4 2007. 537-552 p.
Segl, K.; Roessner, S.; Heiden, U.; Kaufmann, H.: Fusion of spectral and shape features for identification of urban surface cover types using reflective and thermal hyperspectral data. In: ISPRS Journal of Photogrammetry and Remote Sensing, 58, 1-2 2003. 99-112 p.
Roessner, S.; Segl, K.; Heiden, U.; Kaufmann, H.: Automated differentiation of urban surfaces based on airborne hyperspectral imagery. In: IEEE Transactions on Geoscience and Remote Sensing, 39, 7 2001. 1525-1532 p.