Skip to content

Latest commit

 

History

History
22 lines (18 loc) · 1.16 KB

README.md

File metadata and controls

22 lines (18 loc) · 1.16 KB

pyParticlePointTracker

A simple particle filter (also called "sequential montecarlo simulation") that utilizes the farnebäck optical flow and can be extended by additional features. See examples for usage. The farnebäck optical flow restricts the search area in such that you should use fluent videos to track points within. If the jumps are too big for farnebäck to be registrated, then the points will not be moved at all!

Installation

pip install git+https://github.com/Centrasis/pyParticlePointTracker

Importing

from ParticlePointTracker import ParticlePointTracker, Features -> see examples

Predefined Features

  • RGB Histograms
  • HSV Histograms
  • Template matching
  • Grey level coocurence matrix
  • SIFT feature matching

Adding new Features

Just extend the ParticleFeature class to add new features. Feature instances should store their ground truth and current observed values for each point (I just use dicts for that). The calculated error should then be normalized between 0..1. All errors will be normalized by default, but unnormalized error values will lead to a very unbalanced behaviour.