DMCA
A Multi-Time Step Method to Compute Optical Flow with Scientific Priors for Observations of a
BibTeX
@MISC{System_amulti-time,
author = {Fluidic System and Ranil Basnayake and Erik M Bollt},
title = {A Multi-Time Step Method to Compute Optical Flow with Scientific Priors for Observations of a},
year = {}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Abstract Optical flow is a classical problem in computer vision, but the concepts must be adapted for applications to other fields such as fluid mechanics and dynamical systems. Our approaches are based on an inverse problems formalism, considering imposed scientific priors in the form of a cost function that rewards an assumed infinitesimal generator commensurate with assumed physics of the observed density evolution. This leads to a practical and principled approach to analyze an observed dynamical system. Additionally we present here for the first time a new multi-frame version of the functional coupling of multiple images. Following the calculus of variations, this yields a coupled set of Euler-Lagrange PDEs which serve as an assimilation method that inputs video frames as driving terms. The solution of the PDE which follows is the vector field, as designed. Data from an oceanographic system will be highlighted. It is also shown here how these flow fields can be used to analyze mixing and mass transport in the fluid system being imaged. 1
Keyphrases
scientific prior multi-time step method compute optical flow inverse problem formalism vector field abstract optical flow video frame observed density evolution multiple image new multi-frame version dynamical system coupled set driving term computer vision functional coupling flow field fluid mechanic classical problem observed dynamical system assimilation method oceanographic system assumed infinitesimal generator commensurate assumed physic mass transport fluid system first time principled approach cost function euler-lagrange pdes
