class-tracker is a C++ library that implements both an Extended Kalman Filter (EFK) and an Unscented Kalman Filter (UKF), based tracker.
The tracker only works on the position of the object (x,y) to predict not only the new position (x’,y’), but also the velocity v the yaw , and the yaw-rate
. Hence, the state of EKF is:
While the state transition adopted:
It is important to know that the filter expects to receive data in meters and returns:
(x',y')in metersvin m/sin radians
in radians/second
Moreover, it is important to set the correct delta t and the wanted age factor when using the tracker as a library.
sudo apt-get install libeigen3-dev python3-matplotlib libpython3.6
This library also depends upon:
- matplotlib-cpp for visualization
- geotedic_utils for the convertion from GPS to meters
git clone https://github.com/mive93/tracker_CLASS
cd class-tracker
git submodule update --init --recursive
mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE=Release ..
make -j4
Optionally
cmake -DCMAKE_BUILD_TYPE=RelWithDebInfo ..
cmake -DCMAKE_BUILD_TYPE=Debug ..
Matplotlib can be activated (resp. deactivated) with the cmake option -DWITH_MATPLOTLIB=True (resp. -DWITH_MATPLOTLIB=False )
This repository offers a library to exploit the implemented filter, however, there is also a dummy example of the usage of the trackers given by the program tracker. Once the project has been built, it can just be run with:
./tracker
It exploits the file ../data/test_ll.txt in which in each line there is
- frame number
- timestamp
- latitude (GPS)
- longitude (GPS)
and it shows how to convert them into meters using the geodetic_converter.
Once run, it will show the ground-truth (noisy positions) in red and the output of the filter (the prediction of EKF of UKF) in green as in this picture:
This work has been supported by the EU H2020 project CLASS, contract #780622.