Welcome to ParticleTracking’s documentation!
Note
This project is under active development.
ParticleTracking is a repository containing two Python packages for extracting 3D particle coordinate data from the experimental video footage. In these experiments a granular gas (dilute ensemble of macroscopic particles) was observed with a 2-view stereo camera system. The current version of the program is used for experiments with rod-like particles. Each particle is parametrized by 2 endpoint coordinates. ParticleTracking’s goal is to automatically extract the coordinate data and provide a Graphical User Interface (GUI) to correct detection errors manually in an efficient way. Then, the corrected particle data can be tracked and ensemble statistics (average velocity/kinetic energy, local packing fractions, etc.) can be extracted.
The first part is the ParticleDetection package that enables:
training of Mask-RCNN models for detecting particles in stereo-camera images
assignment of particle correspondences between both images of the stereo setup
reconstruction of 3D coordinates and orientations for particles identified in both stereo images
tracking of particles over the course of an experiment
The RodTracker package is a GUI encapsulating the most used functionality of the ParticleDetection package. It enables users to carry out the aforementioned tasks, except for training a model. Additionally, it provides the means to manually correct object localization, 3D assignment and tracking errors in the automatic detection processes.