You are here

Unmanned Aerial System (UAS)

The Research Institute for Nature and Forest (INBO) has a mandate to monitor Natura 2000 habitats in Flanders. Remote sensing by aircraft and satellite has helped to conduct vegetation monitoring more efficiently, but these methods have their limits in flexibility and image resolution.

To monitor those habitats more efficiently, the Gatewing X100, an Unmanned Aerial System (UAS), is deployed. The lightweight aircraft flies on autopilot at high speed and low altitudes, and is equipped with an RGB or Near-Infrared camera. This allows the researchers to photograph at very high resolutions (pixel size up to 5x5 cm) and several times a year, even in windy or cloudy weather. The analysis of the resulting images is facilitated by using object-based image analysis software.

   

Left: Deploying the Gatewing X100 (©INBO) - Right: Natura 2000 habitats in Belgium (Source: Wikipedia and http://natura2000.eea.europa.eu/)

 

The UAS is now operational in the field. Despite the ongoing embargo by the Directoraat-generaal Luchtvaart (DGLV) on issuing permanent flight permissions for unmanned aircrafts, the INBO was able to perform 29 flights in 2015, for which they received temporary permission. Hence, high resolution images of the nature reserves Zwin, Kalmthoutse heide, Eigenbilzen, Averbode and De Liereman could be obtained.

Data processing is now moving to production, using cloud infrastructure (Amazon Web Services). The graphical processing units (GPU) of the machine are used to perform the processing in parallel. The Photoscan software is used to stitch individual images into a large orthorectified image of the entire area, while INBO collaborates with the Flemish Institute for Technological Research (VITO) for segmentation, object recognition and classification. This will lead to derived products such as annotated GIS layers which are important tools to assess the state of the habitats in these reserves.

In 2016, streamlining of the data processing will continue and the generation of these derived products will be included in the automated processing pipeline. However, manual intervention can never be removed completely from the process since the optimization of parameters for segmentation and object recognition cannot be automated.

The results gathered so far will be made publicly available in the course of 2016.