About

OpyFlow : Python package for Optical Flow measurements

Opyflow is a basic image velocimetry tool to simplify your video or frame sequences processing.

It is based on opencv and vtk libraries to detect Good Features to Track (GFT), calculate their displacements by the Lukas Kanade method and interpolate them on a mesh. This method is sometimes called Feature Image Velocimetry or Feature Tracking. It is an alternative to the classical cross-correlation techniques employed in Particle Image Velocimetry (PIV). Compared to this technique, GFT+OpticalFlow may result in better performance when image qualities are poor for velocimetry, i.e. when velocity information on frames is non-uniform.

For flow calculations, the process is mainly inspired by the openCV python sample lktrack.py.

The package also contains some rendering tools built with matplotlib. Velocities can be exported (csv, tecplot, vtk, hdf5).

Author: Gauthier Rousseau

Corresponding e-mail : [email protected]

Quick start

Assuming that you already have an environment with python installed (<=3.9), run the following command on your terminal:

pip install opyf

or from the opyflow repository

python setup.py install

This should install the opyf library and the main dependencies (vtk and opencv) automatically.

If you meet compatibility problems on your system, it is recommended to create an environment via conda installation (see bellow installation with anaconda).

To analyze a frame sequence (png, bmp, jpeg, tiff) you may run the following script:

import opyf
analyzer=opyf.frameSequenceAnalyzer("folder/toward/images")

For a video (mp4, avi, mkv, ... ):

analyzer=opyf.videoAnalyzer("video/file/path")

To perform your first analyze run :

analyzer.extractGoodFeaturesAndDisplacements()

opyf package contains two frames and one video for testing and practicing yourself:

• The two frames were extracted from the frame sequence of the Test case A of the PIV Challenge 2014

When applied to the entire dataset, It can produce the above result (see Test PIV Challenge 2014 - Case A for details on the procedure).

• The video is a bird eye view video of a stream river taken by a drone and from which surface velocities can be extracted (see the following python file for the different possible procedures ).

• A guide to obtain the free surface velocity in a river from two amateur video is also provided :
  Stabilization, orthorectification (brid eye view transformation) and free surface velocity estimation on the Brague river with opyFlow

Contents

This repository is organized as follows:

The setup file:

• setup.py

The package Folder opyf:

• opyf

  • Analyzer.py
  • Track.py
  • Interpolate.py
  • Files.py
  • Filters.py
  • Render.py
  • Tools.py
  • custom_cmap.py (based on Chris Slocum file)

• test

  • Test_case_PIV_Challenge_2014
  • Test_land_slide_youtube_video
  • Test_Navizence
  • Test_Brague_flood

A test on synthetic images is still required.

Installation with anaconda

The package requires python and basic python package: csv, numpy, matplotlib, tqdm, hdf5. Main dependencies are: OpenCV and VTK.

If the pip install opyf command above does not work for you, the simplest way to deal with incompatibilities is using miniconda or anaconda.

When miniconda/anaconda is installed you may create an environment (here called opyfenv). To create the environment type in the terminal:

conda create -n opyfenv python=3.6 vtk opencv matplotlib scipy tqdm (spyder)

This command line will install an environment with python 3.6 and the main dependencies.

You can access to your environment by typing:

source activate opyfenv

From there, you should be able to install opyf with pip install opyf.

For python 3.7 you should use the conda-forge channel:

conda create -n opyfenv python=3.7 vtk opencv matplotlib scipy tqdm (spyder) --channel conda-forge

Tested on: Python version: 3.6 and + VTK : 7.0.1 and + opencv : 3.2 and + numpy: 1.17 and + matplotlib : 2.0.0 and +

Citation

This package has been developed in the course of my PhD at EPFL to study Turbulent flows over rough permeable beds. Outputs are visible in the manuscript as well as in this Video where paraview animations have been rendered thanks to opyf outputs.

@PhdThesis{rousseau2019turbulent, title={Turbulent flows over rough permeable beds in mountain rivers: Experimental insights and modeling}, author={Rousseau, Gauthier}, year={2019}, institution={EPFL} }

An article published in Experiments in Fluids is available in Open Access : Scanning PIV of turbulent flows over and through rough porous beds using refractive index matching

Follow this link to directly access to the annex of the article presenting the opyf algorithms and tests.

@article{rousseau2020scanning, title={Scanning PIV of turbulent flows over and through rough porous beds using refractive index matching}, author={Rousseau, Gauthier and Ancey, Christophe}, journal={Experiments in Fluids}, volume={61}, number={8}, pages={1--24}, year={2020}, publisher={Springer} }

Contributors : Hugo Rousseau, Mohamed Nadeem, LHE team and others

Credits for UAV video : Bob de Graffenried

Credits for 2019 flood video on Brague River (Biot, French Riviera) : Pierre Brigode