Surface segmentation technique for tomographic PIV: Adaptive surface and iterative 2-D interrogation

Tomographic PIV has become a widely used 3-D flow measurement technique. It utilizes projected images recorded by cameras from different views and iterative algebraic reconstruction technique to reconstruct the intensity distribution of a measured volume, and computes the 3D3C velocity field by 3-D cross-correlation. Surface segmentation technique was proposed by Adrian (2011) accelerates the reconstruction process by extracts from the cloud of particle images only those images originating from particles that lay on a mathematically prescribed surface(s). 2-D cross-correlations are performed on orthogonal stacks of surfaces to compute the 3D3C velocity field, which further reduce the computational cost compared to 3-D crosscorrelation. The method is not as effective as the MART techniqueIn this study, we investigate the reconstruction of adaptive surfaces. Those surfaces are chosen so that the particle out-of-plane motion is minimized and 2-D interrogation works well. Numerical assessments are performed for simulated vortex ring and Couette flow. Furthermore, iterative 2-D interrogation on orthogonal stacks of surfaces combined with volume deformation is proposed to refine the velocity field and save the computational cost. The test on simulated vortex ring shows the global r.m.s error can be reduced below 0.2 voxels (See Table 1). non-iterative 0.2178 0.2100 0.2996