TracTrac is an open-source Matlab/Python implementation of a robust and efficient object tracking algorithm capable of simultaneously tracking several thousands of objects in very short time. Originally developed as an alternative to particle image velocimetry algorithms for estimating fluid flow velocities, its versatility and robustness makes it relevant to many other dynamic sceneries encountered in geophysics such as granular flows and drone videography. In this article, the structure of the algorithm is detailed and its capacity to resolve strongly variable and intermittent object motions is tested against three examples of geophysical interest.
TracTrac is fast (more than 10000 points tracked per second) and accurate (up to 0.01 pixel resolution), forming thus a good concurrent to the state-of-the art PIV algorithms. It allows to track anything that moves: birds, ants, grains, water flows... It runs on Python (v2.7/3.6/3.8 with OpenCV2) or Matlab (>2014a with Computer Vision and Statistics toolbox). Give it a try !
**From version v3.0, development of TracTrac will be only continued for Python. Python environment now offers graphical outputs and live adaptation of parameters via the text parameter file.
TracTrac has been tested in the following reference:
Heyman J., TracTrac: A fast multi-object tracking algorithm for motion estimation, Computers & Geosciences, Volume 128, 2019, Pages 11-18,doi: 10.1016/j.cageo.2019.03.007
The article is available in my personal webpage
Contact: joris.heyman@univ-rennes1.fr
- Installation Download and extract the Git repository. Make sure you have a working installation of python2.7 or python3.xx. Install the extra packages scipy, opencv-python, imutils, h5py, parse via the terminal :
pip3 install scipy opencv-python imutils h5py parse vispy pyqt5 --user
- In the Python folder, run test case in GUI, plot and save average velocities with
python3 tractrac_GUI.py -a
2.1) In the Python folder, run test case in command line, plot and save average velocities with
python3 tractrac.py -a
- Get help on available commands with :
python3 tractrac.py --help
- Other sample cases may be run as
python3 tractrac.py -f "../Sample_videos/videotest.avi"
Run live-treatment by webcam acquisition by
python3 tractrac.py -f "0"
Tracking parameters can be changed (before and during processing) in the txt file "*_par.txt" which is automatically created in the video folder. Precise meaning of parameters are given in the publication Heyman (2019) .
-
Installation Download and extract the Git repository. Make sure you have Matlab version >2014a with Image Processing Toolbox and Statistics and Machine Learning Toolbox installed.
-
In the Matlab folder, run TracTrac Graphical User Interface with
tractrac
A basic level of output is obtained with the average velocity maps saved as 32bit tiff image files (option -a in python). The images can then be post-treated with an image software such as Fiji/ImageJ. Image filenames contain extremum field values "[min,max]" so that images can be scaled appropriately.
A second level of post-processing is available in the Matlab GUI, with specific plotting functions on a defined grid. A post-processing binary matlab file is generated for further used. It contains cell arrays with trajectories mapped on the defined grid.
If these two levels are not sufficient, the raw trajectories can also be saved in ASCII or binary files (hdf5) for further use.