FLUID European Project

stickshrivelΜηχανική

24 Οκτ 2013 (πριν από 4 χρόνια και 2 μήνες)

78 εμφανίσεις

AMI GROUP

University of Las Palmas de Gran Canaria

FLUID European Project

Activity report of the first year and futur plans

December 2005

FLUID Specific Targeted Research Project
-

Las Palmas Meeting

2

Contents



AMI group short presentation


Activities in WP2 and future plans


Activities in WP3 and future plans


Activities in WP4 and future plans


Project risk analysis



December 2005

FLUID Specific Targeted Research Project
-

Las Palmas Meeting

3

AMI group people


Aleman Miguel (Permanent position at ULPGC)


Alvarez Luis (Permanent position at ULPGC)


Castaño Carlos (Engineer, financed by FLUID project since
october 1, 2005)


García Miguel (Engineer, financed by FLUID project since
november 1, 2005)


González Esther (Permanent position at ULPGC)


Krissian Karl (PostDoc position, financed by FLUID project
since december, 2005


Mazorra Luis (Permanent position at ULPGC)


Salgado Agustín (University grant holder, financed by the
ULPGC)


Sánchez Javier (Permanent position at ULPGC)

December 2005

FLUID Specific Targeted Research Project
-

Las Palmas Meeting

4

AMI group background


Mathematical Analysis


Multiscale Analysis


Partial differential equations


Nonlinear filtering


Optic flow estimation techniques


Shape Analysis


3D geometry reconstruction from 2 or multiple
views


No previous experience in fluid image sequence

December 2005

FLUID Specific Targeted Research Project
-

Las Palmas Meeting

5

WP 2: Physical models and early image
processing. Activities


Analysis of optic flow estimation techniques


Multiscale approach using structure tensor
information at different scales


PDE based approach


Correlation methods


Report 1:
Optic flow estimation in fluid images I


Report 2:
A Methodology to Compare Optic
Flow Estimation methods

December 2005

FLUID Specific Targeted Research Project
-

Las Palmas Meeting

6

WP 2: Physical models and early image
processing. Future plans


To continuous to work on 2D multiscale analysis


To combine correlation based sparse methods and
dense variational approach

December 2005

FLUID Specific Targeted Research Project
-

Las Palmas Meeting

7

WP 3: 2D Fluid motion analysis.
Activities (Questions)


Fluid mechanics evolution models. PDE equations ?


Navier
-
Stokes ? Parameters?


Vorticity equation ?


Pressure equation ?


Coriolis force ?


Hodge decomposition ?


Can we hope (by adding constraints) to solve PDE fluid
mechanics evolution models in a 3D neighbourhood of a 2D
layer where we have a motion estimation ?

December 2005

FLUID Specific Targeted Research Project
-

Las Palmas Meeting

8

WP 3: 2D Fluid motion analysis. Vortex
models I


Fluid mechanics evolution models. Vortex models

December 2005

FLUID Specific Targeted Research Project
-

Las Palmas Meeting

9

WP 3: 2D Fluid motion analysis. Vortex
models II


Fluid mechanics evolution models. Vortex models

December 2005

FLUID Specific Targeted Research Project
-

Las Palmas Meeting

10

WP 4: 3D Fluid motion estimation.
Satellite sequences (layer segmentation)


Compute 2D flow estimation.


Transform of flow information from pixels to physical
measures (motion in Km)


Compute the height of each point using infrared temperature
intensity value.


Compute a segmentation of the infrared temperature
channel


Implementation of a 3D visualization tool for the layers


Tracking regions using a motion model


A robust tracking method could deal with layer occlusions.



December 2005

FLUID Specific Targeted Research Project
-

Las Palmas Meeting

11

WP 4: 3D Fluid motion estimation. PIV
stereoscopic motion estimation


Calibration. Alignment between the light sheet and the
calibration plate


PIV 3D motion estimation using 2 or multiple views.

December 2005

FLUID Specific Targeted Research Project
-

Las Palmas Meeting

12

Project risk analysis


We need to design tractable applied physical models. The interplay
between 2D image measurements and 3D dynamical system is
perhaps too complex.


We need to design a multi
-
layer coupling model (too complex).


Computer vision groups of the project need to strongly interact
with the other partners.


We need to define clearly a way to share results and a validation
strategy.


We need to identify specific tasks in the project where different
partners work together.


We need to convince Bruxelles people that everything is fine.