Context and objectivesThe main application discipline of the research project is “terrestrial ecology”. Strong links exist to “environment”, “agriculture” and “forestry”.
The research aims to develop improved data processing algorithms that take full advantage of the spectral information captured by hyperspectral imagers in order to provide accurate and reliable vegetation maps (qualitative and quantitative characteristics) required by (public) decision-makers for the sustainable management of coastal wetlands.
The main objective of the research is to further develop object-based approaches for hyperspectral image analysis. The aim is to facilitate a sustainable management of wetland ecosystems. The algorithms to develop should allow for segmenting hyperspectral data cubes into image objects with clear object boundaries, to classify the image-objects into meaningful categories (i.e. vegetation types), and to extract and assign relevant biophysical (leaf area index, percent cover) and biochemical vegetation properties (leaf chlorophyll/nitrogen content) to those units.
Expected scientific results
From the research project important outcomes are expected:
- Increased scientific knowledge & expertise in the fields of imaging spectroscopy (IS), hyperspectral image classification and segmentation, inversion of radiative transfer models (RTM) for biophysical and biochemical parameter retrieval, genetic algorithms (GA) for feature selection, artificial neural networks (NN) for RTM inversion, object-based approaches for EO data to reduce the ill-posed nature of model inversions
- Development of new & improved algorithms for hyperspectral data processing: image segmentation and classification and retrieval of biophysical and biochemical variables
- Accurate and reliable thematic maps of Schiermonnikoog Island for sustainable environmental management of: distribution of vegetation (saltmarsh vegetation types), distribution of biophysical and biochemical variables (LAI, percent cover, leaf chlorophyll/nitrogen content), change maps of vegetation types and biophysical/biochemical variables between 1999 and 2005
- Establishment of international research networks for hyperspectral remote sensing including VITO, University of Gent, WUR, RWS and ITC
- Capacity building for PhD, MSc and Erasmus Mundus students involved in the project and Scientific (peer-reviewed) publications
|Project leader(s):||UGent - Remote Sensing / Spatial Analysis lab (REMOSA)|