Holograph g phic Particle Image Ve elocimetry Michael Barry, Alish ha Schor, Anna Shih Mayy 6, 2009 Motivation Image removed due to copyright restrictions. Please see Fig. 7 in Meng, Hui, et al. "Holographic Particle Image Velocimetry." Measurement Science and Technology 15 (2004): 673-685 673 685. Near field of a laminar je jet image from Meng et. al, 2000 Turbulent flow in a square duc duct Image from http://www.me.jhu.edu/meneveau/gallery.html Courtesy of Bo Tao. Used with permission. Holography - Basic B Principles Recording g Reconstruction Image removed due to copyright restrictions. Please see Fig. 1 in Meng, g, Hui,, et al. "Holographic g p Particle Image g Velocimetry." y Measurement Science and Techn nology 15 (2004): 673-685. 3d information to 2d representation p Reconstruct at discrete z depths image from Meng et. al, 2000 Basic Illumin nation Types In-Line • ‘in-line’ reference beam • simpler p set up p • small diameter particles Image removed due to copyright restrictions. Please see Fig. 6a,b in Meng, Hui, et al. "Holographic Particle Image Velocimetry." Measurement Science and Technology 15 (2004): 673-685. Off A is Off-Axis •obliquely incident reference beam • real and virtual images separate during reconstruction reconstructio • higher seeding densities • for digital holography: i id t angle incident l < 2° 2° image from Meng et. al, 2000 Experimental Design Considerations • Recording distance e (Zhang 2006) - If large z: low pas ss filter - If small z: limited FOV, aliasing • Particle concentratiion (Kim, (Kim 2008) - Speckle noise vs. flow information recovery • CCD: size size, pixel sizze Experimen ntal Setup Camera Flow tank Iris Spatial filter Neutral density f filter Diode laser Collimating lens Raw Image Cylinderr Flow Direction Crop initial image to square Reconstruc ction Code (Convolutio on Method) Take the Fourier Transform Reconstructed R t t d image Multiply this by the Fresnel “diffraction kernel” Take the inverse Fourier Transform exp{-iπλz (u2+v2)} note the dependence on z dept depth Reconstruc cted cted Image Out of focus particle In focu focus particle Reconstructtion tion In Depth p Particle D Detection t1 t2 tn Input: Time series of raw holograms 1. Image Preproc cessing and Reconstrruction: • Subtract S bt t b backg kground d • Reconstruction n • Threshold • Convert to B&W W 2. Locate Particles in x 2 x-y: -y: • bwlabel (MATLAB) • Bounding box and a centroid of contiguous white zones 3. Eliminate invallid particles 3 • Outside of expeccted size • Double-counting a single particle Output: matrix of labeled particles with x-y position of centers and area for each zz-plane plane Lab el 1 2 … n x y are a Particle Dete Detection ction (cont’d) (cont d) z1 z2 zn x-y info Input: xx-y -y coordinate matrices at each z-plane 1. Load output of o previous step p • zz-plane plane is determined by step size 2. Eliminate dou ubles: • Particles may appear in focus in multiple e planes • Choose based on o minimum radiuss Output: matrix of labeled particles with x-y-z position of centers and radius for each time step Label x y z radius 1 2 … n t Result: 3 3D D Flow Flow Calculating V Velocity Maps Map • HPIV typically uses 3 3D coordinates as dat data - Computing cost of images is high (96 GB) • Tradeoffs depending g on complexity of flow, mean velocit velocity - Correlation: complex flows, dense particle seeding - Particle tracking: lo ow velocity flow flows, sparse seeding (re educes speckle noise) Velocity Maps Map by Correlation Correlatio • Correlation (Pu & Meng, M 2000) noise 0 1 2 t1 1 0 2 4 3 3 4 t2 Two spatially shifted images (time sequence sequence) t1t 1 10 1 0 1 1 0 0 0 0 1 3 1 33 3 3 3 2 2 2 2 4 4 2 2 4 4 4 4 t2t 2 t2t 2 Mean Meandisplacement displacementvector vector • Shift images according to o mean velocity vector • Within discrete interrogation windows, pair particles based on closest distance e, group deformation threshold Velocity Map ps by Particle Trac T cking ki • Pair particles by miinimizing a cost function (Stellmacher & Obe ermayer 2000) • Utilize Utili kknowledge l d about ab t flow: fl rr r Yi = AX AXi + t • Cost function J K E(m,t, A) = ∑ ∑ m jk Y j − t − AX k j =1 k =1 2 J K j =1 k =1 + g(A) + α ∑ m j (K +1) + α ∑ m( J +1)k re egularizer” egularizer Yk “re Cost for unmatched particles • Maximize expectatiion Velocity Ma Maps - Others • Genetic algorithm p pa article pairing (Shen (Sheng and Meng 1998) - Treat T t pairings ii lik like chromosomes ch - Choose “most fit” pairs p to p propogate p g • Optical correlation (C Coupland and Halliwell 1997) - Expand 2D autocorrrelation to 3D - Can utilize FFT Velocity y Maps Map Image from http://www.me.jhu.edu/lefd/hpivv/3d_vec.pdf Courtesy of Bo Tao. Used with permission. • Questions? Bibliog graphy • Arroyo M; Hinsch K (2008) Recent developments v of PIV towards 3d measurements. Topics Appl. Physics (112) 127-154 • Hinsch KD (2002) Holographic image velocimetry. Meas. Sci. Technol. (13) R61R72 • Kim S; Lee S (2008) ( ) Effect of particle numb ber density y in in-line digital g holographic g particle velocimetry (44) 623-631 • Meng H; Pan G; Pu Y; Woodward S (2 2004) Holographic particle image velocimetry: from film to digital recordiing. Meas. Sci. Technol. (15) 673-685 • Pu Y., Meng H. (2000) An advanced off-axis o holographic partcile image velocimetry (HPIV) system). Exp in F luids (29) 184-197 • Pu Y; Song X; Meng H (2000) Off Off-axis axiss holographic particle image velocimetry fo for diagnosing particulate flows. Exp in Fluids F [Suppl.] S117-S128 • Sheng, J., Meng, H. (1998) A genetic algorithm particle pairing technique for 3D g p p particle image g velocimetry. y Exp p in Fluids velocityy field extraction in holographic (25) 461-473 • Stellmacher, M; Obermayer, K. (2000)) A new particle tracking algorithm based on deterministic annealing and alternative e distance measures. Exp in Fluids (28) 506518 • Zhang Y; Shen G; Schroder A (2006) Influe ence of some recording parameters on digital holographic particle image velocimetry. Opttical Engineering (45) 075801 MIT OpenCourseWare http://ocw.mit.edu 2.71 / 2.710 Optics Spring 2009 For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.