Measuring Drone Trajectory using Total Stations with Visual Tracking [IPA '20 @ ETH]
This repository is for the IPA project Measuring Drone Trajectory using Total Station with Visual Tracking at ETH IGP.
Student: Yue Pan
Supervisor: Dr. Cenek Albl, et al.
In this project, we managed to construct a visual drone tracking and positioning dataset collected by a multi-sensor system, including a total station, on-board sensor kits, and an ad-hoc network of cameras.
By leveraging high accuracy total station measurements and sensor fusion techniques such as the extended Kalman filter, the absolute positioning accuracy for the drones body center in the local frame can be better than one centimeter. By employing a radio-synchronized network of audio triggers, we can recognize the triggering pattern from the audio signal of each video, thus accomplishing the synchronization among the videos with sub-frame accuracy. The overall synchronization time delay of the system is about 10 milliseconds after aligning all the measurements timestamp to the referenced PC time.
Drone visual tracking and trajectory reconstruction dataset [link to the dataset]
The codes are written in Matlab (R2019b).
Leica total station with GeoCOM
Drone assembled with Pixhawk and Leica 360 prism
Radio-synchronized network of audio triggers
Run codes\measure\tps_geocom\main_tps.m
to realize total station tracking via GeoCOM.
Run codes\measure\radio_sync\audio_trigger_auto.m
to launch the radio-synchronized audio triggering system for joint synchronization.
Run codes\postprocess\pose\drone_data_loader.m
to load the drone log file in Pixhawk format.
Run codes\postprocess\pose\get_tran_lt.m
to accomplish resection (transformation from the total station frame to local ENU frame).
Run codes\postprocess\pose\get_drone_pose.m
to estimate the drone's pose in local ENU frame using total station and onboard sensors' measurements.
Drone tracking and trajectory reconstruction dataset
Reconstruction of 3D flight trajectories from ad-hoc camera networks (IROS '20)[Codes]
Special thanks to Prof. Dr. Konrad Schindler, Dr. Cenek Albl, Dr. Jemil Butt, Andreas Baumann-Ouyang, Alexander Wolf, Thomas Posur, Tom Manu, Usvyatsov Mikhail and Mudathir Awadaljeed from ETH Zurich Institute of Geodesy and Photogrammetry for the supervision, advice and help during the project.