PEACHZ - PROBA-V Evaluation of Accuracy Characteristics in Heterogeneous Zones

Context and objectives

The characteristics of PROBA-V images will be strongly influenced by the compression algorithms applied to them prior to transmission.  We propose to carry out an investigation into the impact of these compression algorithms on mapping accuracy from simulated images.  We intend to evaluate the utility of the images for land cover mapping in relation to the different levels of spatial heterogeneity, to determine if the lossy compression of the images results in a loss of accuracy for regions where spatial heterogeneity is high.  The objectives of this work include:

  1. Production of simulated PROBA-V imagery based on the supersampling of existing Landsat TM imagery.
  2. Measurement of spatial heterogeneity characteristics (e.g. texture, wavelet decomposition) across simulated imagery and mapping land cover from simulated PROBA-V, SPOT-VEGETATION and original Landsat imagery.
  3. Evaluation of the impact of compression and spatial resolution of simulated PROBA-V imagery on mapping accuracy across different levels of spatial heterogeneity.
  4. Identification and evaluation of methods for reducing the impact of image compression on mapping accuracy.

Project outcome

Expected scientific results

  1. Dataset comprising Landsat TM and SPOT-VEGETATION scenes for approximately 30 sites globally.
  2. A dataset of simulated PROBA-V scenes, corresponding to the selected study sites.
  3. Spatial characterisation of SPOT-VEGETATION, Landsat TM and simulated PROBA-V scenes.
  4. Land cover maps for each scene (3 different sensor types & 2 classification systems = 6 LC maps per study site).
  5. Statistical characterisation of simulated PROBA-V scenes in relation to different landscape characteristics.
  6. Reporting on methods for restoring or improving land cover mapping capability to compressed PROBA-V scenes.
  7. Intermediate project report, delivered six months from start date.
  8. Final project report, delivered twelve months from start date.
  9. 2 peer-reviewed journal papers and 1 conference paper.