Reconstruction of instantaneous 3D flow density fields by a new direct regularized 3DBOS method

We present a new numerical method for 3D reconstruction of instantaneous density volume from Background Oriented Schlieren (BOS) measurements, with a validation on a dedicated flexible experimental 3DBOS bench. In contrast with previous works, we use a one-step formulation where density is directly estimated from measured deviation fields without the intermediate step of density gradient reconstruction. Regularization techniques are implemented to deal with the ill-posed problem encountered. The resulting high dimensional optimization is conducted by conjugate gradients (CG) techniques with an explicit step computation. A parallel algorithm, implemented on Graphics Processing Unit (GPU), helps to speed up the calculation. The resulting software is validated on a dedicated BOS3D experimental facility which has been built to study various BOS settings. Results illustrate the potential of the method for flow characterization and measurement in real-world conditions.

[1]  Daniel Gross,et al.  GPU-Based Volume Reconstruction from Very Few Arbitrarily Aligned X-Ray Images , 2009, SIAM J. Sci. Comput..

[2]  Guy Le Besnerais,et al.  Multi-camera calibration for 3DBOS , 2014 .

[3]  Ivo Ihrke,et al.  Reconstruction and rendering of time varying natural phenomena , 2007 .

[4]  Markus Raffel,et al.  Background oriented stereoscopic schlieren (BOSS) for full-scale helicopter vortex characterization , 2000 .

[5]  F. Champagnat,et al.  3D RECONSTRUCTION OF THE DENSITY FIELD OF A JET USING SYNTHETIC BOS IMAGES , 2012 .

[6]  F. Champagnat,et al.  Fast and accurate PIV computation using highly parallel iterative correlation maximization , 2011 .

[7]  Hans-Peter Seidel,et al.  Time-resolved 3d capture of non-stationary gas flows , 2008, SIGGRAPH Asia '08.

[8]  J. Idier,et al.  Convergence of Conjugate Gradient Methods with a Closed-Form Stepsize Formula , 2008 .

[9]  Per Christian Hansen,et al.  Analysis of Discrete Ill-Posed Problems by Means of the L-Curve , 1992, SIAM Rev..

[10]  Aurélien Plyer,et al.  Massively parallel Lucas Kanade optical flow for real-time video processing applications , 2014, Journal of Real-Time Image Processing.

[11]  Ali Mohammad-Djafari,et al.  GPU implementation of a 3D bayesian CT algorithm and its application on real foam reconstruction , 2010 .

[12]  Marcus A. Magnor,et al.  Image-based tomographic reconstruction of flames , 2004, SIGGRAPH '04.

[13]  F. Leopold,et al.  Reconstruction of the density field using the Colored Background Oriented Schlieren Technique (CBOS) , 2012 .