Skip to content
Andreas Bircher edited this page Nov 30, 2015 · 9 revisions

Experimental results

The algorithm is actively used within the framework of inspection operations using both rotorcraft-type as well as fixed-wing aerial robots. More specifically, the Firefly MAV platform and the AtlantikSolar UAV were used - both equipped with the VI-Sensor. The platforms are shown in the image below:

The following datasets are available:

The video below presents the recorded flight regarding the inspection of the ETH Polyterrasse:

Inspection of the ETH Polyterrasse

All datasets contain dense point clouds derived using the proposed autonomous inspection planner for the flight path, a camera system and the Pix4D software for post-processing of the pose-annotated images. The point clouds are saved in *.ply format and you may load them with any relevant software (e.g. MeshLab). Furthermore, for the case of the Marche-en-Famenne inspection using the AtlantikSolar UAV equipped with the Sony HDR-AS100VW camera orthophoto data along with Google Earth KML files are provided.

The employed UAVs:

  • AtlantikSolar, a solar-powered long-endurance aerial vehicle designed by our lab and equipped with a monocoluar version of the VI-Sensor. Find more info here.
  • Firefly, a hexacopter vehicle from Ascending Technologies and equipeed with advanced functionalities by our lab along with a VI-Sensor module. Find more info here.

The Camera systems used:

  • The Visual Inertial Sensor (VI-Sensor) developed by ASL and Skybotix that provides tightly aligned Visual frames with IMU information. Find more info here.
  • A modified version of the VI-Sensor onboard AtlantikSolar
  • A Sony HDR-AS100VW GPS-tagged action cam

Example images

  • Below you find example images from the monocular version of the VI-Sensor on-board AtlantikSolar, o f the Sony HDR-AS100VW on-board the same UAV and of the VI-Sensor on-board Firefly.

How to load the Point-clouds:

Loading the point-clouds provided here is a meaningful way to assess the quality of the inspection data. Indeed there exist multiple software solutions that allow processing point-clouds. Among them, Meshlab is a free-of-charge software available for all major operating systems. Download this software and use the File>Import Mesh functionality to load the point-clouds provided here in ASCII-formatted *.PLY form. Below indicative images from the point-clouds.

How to load the Google-tiles:

For the case of the fixed-wing UAV inspection using the Sony HDR-AS100VW nadir-looking camera, google tiles are provided that can be loaded using Google Earth. Run the software and then use the File>Open functionality to load the “Marche_en_Famenne_AtlantikSolar_SonyHDRAS100VW_Google.kml” file. Note that close to the edges, accuracy is lost due to lack of information. Below an example with the google tiles loaded and an inspection path plotted on-top.

Clone this wiki locally