08 - Dynamic Experiments

Question 1

Big data

Answer 1

• volume
• velocity
• variety

Question 2

Object tracking

Answer 2

• nearest neighbor method
• challenges = long series, random (dis)appearances, jitter/motion noise
• improve with acquisition-based param (acquisition rate, resolution) + xpmtl param (setup, vibrations, T°, polydispersity, mixture)
• extending nearest neighbor
• – bijective requirement
• – prior/expected movement
• – adaptative movt
• – multiple hyp testing
• limits: merge/split ptcls
• voxel-based approach = std img correlation

Question 3

Registration

Answer 3

• 1st step = align on the same coordinate system
• process: transform op°, interpolator, metric, optimizer
• ITK and SimpleITK (registration for medical imaging)
• qtfying deformation = strains
• – volumetric = isotropic (size, scale)
• – deviatoric = anisotropic (proportions)
• 2-pt correlation = characterize shape each time step and compare
• feature points
• – corner detection
• – use descriptors (BRIEF, not the best; ORB, orientation dpdt)