With reference to the weaknesses of current remote sensing methods in habitat reporting, the objective of this project to is to enhance the state-of-the-art classification framework. In this project, we intend to create a transferable platform which integrates novel and advanced remote sensing methodologies that are developed specifically for operational habitat reportage.

Previous efforts have been put in a classification framework with emphasis on mapping vegetation species based only on their spectral signature. In this work, we will extend this classification framework with the introduction of spatial aspects to the classification. Resulting features can complement spectral features, or can compensate a lack of detailed spectral information.

Remote sensing of vegetation is highly complex. Numerous parameters influence the reflectance signature of the canopy. Not only atmosphere, but also biophysical parameters such as leaf area index, leaf angle distribution and viewing geometry have their impact. A sensitivity analysis of those parameters will be performed and a multiresolution approach will be introduced to retrieve more robust features. We also want to better understand the impact of scale, resolution and time on classification of ecotopes, which might lead to more costeffective approaches. We want to further explore the link between different scales 1) leaf - top of canopy - top of atmosphere and 2) airborne imagery and coarser satellite data (e.g. Chris-Proba 18m).

Third, this project will use the latest state-of-the-art super resolution (SR) image reconstruction algorithms to narrow the gap of spatial resolutions between airborne and spaceborne hyperspectral data. SR algorithms are devised to obtain a high resolution image from multiple lower resolution images. These techniques have existed for decades, but have gained interest due to the generalization and improvement of digital imaging in different fields such as medical imaging, satellite imaging, video surveillance, etc. We will investigate the potential of SR algorithms in hyperspectral image and the information gain of the SR images for ecotope classifications.

To improve classification accuracies and to strengthen an operational-oriented classification chain, our fourth objective is to investigate the operational potential of ensemble  classifiers in terms of stability, accuracy, ease of use, and computing costs.

The fifth objective is to introduce structural analysis to use the heterogeneity of vegetation types for better identification and to establish the link with the degree of development of habitat types.

Our last objective is to integrate and validate our developed methodologies using the Belgium Biological Valuation Map and selected habitats which are prioritized under Natura2000.