Vision & Laser Datasets From A Heterogeneous UAV Fleet

T. Hinzmann, T. Stastny, G. Conte, P. Doherty, P. Rudol, M. Wzorek, E. Galceran, R. Siegwart, and I. Gilitschenski, Collaborative 3D Reconstruction using Heterogeneous UAVs: System and Experiments. In International Symposium on Experimental Robotics (ISER), 2016, Tokyo, Japan.PDFBibTex

The datasets were collected in Italy and Sweden in the context of a joint mission with the Autonomous Systems Lab, ETH Zurich and the Department of Computer and Information Science, Linköping University; funded by the FP7 EU project SHERPA.

Figure 1: Point cloud generated by the SICK laser scanner from the Rotary-Wing UAV (left) and the vision point cloud generated by the fixed-wing UAV (right).

Employed Platforms & Exteroceptive Sensors

Yamaha RMAX (Rotary-Wing UAV): The Yamaha RMAX helicopter has a rotor diameter of 3.1 m, a maximum take-off weight of 94 kg and a payload capability of about 30 kg. The platform is capable of fully autonomous navigation, including take-off and landing. The basic sensor suite used for autonomous navigation includes a fiber optic tri-axial gyro system and a tri-axial accelerometer system, a RTK GNSS positioning system and an infrared altimeter used for automatic landing.

Techpod (Fixed-Wing UAV): Small unmanned research plane with a classic T-tail configuration, is equipped with one propeller, has a wingspan of 2.60 m and a nominal speed of around 12 m/s. The sensor and processing unit as well as the PX4 auto-pilot are located inside the modified fuselage and allow autonomous mission execution such as GPS waypoint following.

Platform Exteroceptive Sensors
RMAX SICK LMS511 PRO 2D laser scanner, maximum range: 80m, maximum scanning FoV: 190°
Techpod Sony ActionCam HDR-AS100V, RGB, rolling shutter, 13 MP; Aptina MT9V034, grayscale, global shutter, 0.3 MP

Figure 2: The UAV platforms employed for the cooperative scanning missions.

Available Datasets

Dataset Name Date (Session) Location Sensor Terrain Preview Download
laser_motala_1 2015 (I) Motala, Sweden SICK House area png laz
laser_motala_2 2015 (I) Motala, Sweden SICK House area png laz
laser_motala_3 2015 (I) Motala, Sweden SICK House area png laz
laser_motala_4 Feb 2016 (II) Motala, Sweden SICK House area png laz xyz
laser_motala_5 Feb 2016 (II) Motala, Sweden SICK House area png laz
laser_motala_6 Feb 2016 (II) Motala, Sweden SICK House area png laz
vision_motala_1 Mar 2016 (III) Motala, Sweden Sony Complete overview png laz
vision_motala_2 Mar 2016 (III) Motala, Sweden Sony House area, cropped png xyz
laser_motala_7 Mar 2016 (III) Motala, Sweden SICK House area png xyz
laser_motala_8 Mar 2016 (III) Motala, Sweden SICK House area png xyz
laser_motala_9 Mar 2016 (III) Motala, Sweden SICK House area png xyz
laser_motala_10 Mar 2016 (III) Motala, Sweden SICK House area png laz
vision_isollaz_1 Apr. 2016 (VI) Isollaz, Italy Aptina Hillside, groves, stone steps, pathspng laz
laser_isollaz_1 Apr. 2016 (VI) Isollaz, Italy SICK Grove, stone steps, pathpng laz
laser_isollaz_2 Apr. 2016 (VI) Isollaz, Italy SICK Grove, stone steps, pathpng laz
laser_isollaz_3 Apr. 2016 (VI) Isollaz, Italy SICK Grove, stone steps, pathpng laz
laser_isollaz_4 Apr. 2016 (VI) Isollaz, Italy SICK Grove, stone steps, pathpng laz

Figure 3: Test site near Motala, Sweden (lat/lon: 58.495043/15.102358) and Isollaz, Italy (lat/lon: 45.686922,/7.717967)

Useful Tools

For instance LAStools (in combination with wine on Linux) can be used to view, convert and crop point clouds in .las, .laz and .xyz format.

# view
wine LAStools/bin/lasview pointcloud.las

# convert 
wine LAStools/bin/las2las -i input.las -o output.xyz

# crop
wine LAStools/bin/las2las -i input.xyz -o output.xyz -keep_xy x_min y_min x_max y_max

Source Code

The source code used for aligning a laser and a vision point cloud is available here: Source Code: Robust Point Cloud Registration. The repository contains wrappers for ICP, GICP, NDT as well as the source code for IPDA which is a robust point cloud registration method that uses one-to-many probabilistic data associations (“Robust ICP”).

Figure 4: Misaligned laser point cloud (green) and vision point cloud (colored).

Figure 5: Aligned laser point cloud (green) and vision point cloud (colored) using Robust Point Cloud Registration

Additional Information

In case of questions, please contact the authors at hitimo[at]ethz.ch. Many thanks to our great partner from Linköping University.


You are here: homeiser2016
Valid CSS Driven by DokuWiki Valid XHTML 1.0