HYPERFOREST - Advanced airborne hyperspectral remote sensing to support forest management

You are here

Context and objectives

The HyperForest project – a consortium of K.U.Leuven, UGent, VITO, GLI, INBO, RSL - aims at providing foresters with detailed spatial explicit data on forest vitality, species composition and stand diversity based on airborne hyperspectral and LiDAR data.

The complex nature of hyperspectral data sets urge producers to set up a complete imagery pre- processing chain to perform standard corrections for radiometric, geometric and atmospheric effects which might corrupt the data. Moreover, bidirectional effects caused by the heterogeneous character of terrestrial targets, such as forests which have pronounced vegetation structures, are affecting the captured hyperspectral signal. This project aims at:

  • developing an advanced airborne hyperspectral imagery pre-processing chain (e.g. APEX) that considers vegetation structure (bidirectional) effects of the reflected signal,
  • delivering of a robust methodology to extract optimized vegetation indices quantifying forest diversity from this pre-processed imagery.
  • organizing intensive interactions with end-users by considering their feedback facilitating the supply of tuned and more end-user oriented forest thematic products.

Basically, with this project we want to identify forest canopies components that contribute the most to the captured reflectance values of airborne sensors.

Project outcome

  • Tree vitality was detected using HS indices derived from hyperspectral airborne remote sensing data. However, a HS-index performing well in a particular test location, is not necessarily useful in another study area;
  • A method was developed, based on PCA and cross-validation, for fusion of HS data and LiDAR height profile information which resulted in improved tree species mapping;
  • A novel algorithm for hyperspectral image denoising was developed which led to more stable tree species classification results; higher accuracy levels, reduced feature selection processing time and reduced number of selected bands subsequently reducing the classification time of the SVM classifier;
  • A filtering algorithm to smooth APEX data was implemented in an operational processing chain;
  • The inversion procedure was applied to entire APEX images that had been acquired over the project test sites. Images of leaf area index and total chlorophyll were produced and partially validated;
  • The radiative transfer models SLC and DART were implemented to the programming environment to allow for a fast computation of image data;
  • Following 3D structural information products were provided for the three test sites: voxel layers, single-tree geometry and layers of vegetation density parameters such as LAI and canopy closure;
  • The 3D arrangement of forest canopy elements was reconstructed;
  • The contribution of the vegetation structure to the hyperspectral reflectance was evaluated through simulations on the physically based ray tracer.