Evaluation of a wavelet-based optical flow method for planar velocimetry using scalar fields

A wavelet-based optical flow velocimetry (wOFV) method is evaluated in the context of determining flow velocity from scalar images. A hierarchy of data sets are used to provide an assessment of the scalar-based methodology, including synthetic image pairs of passive scalar fields from 2D and 3D DNS sets and experimental data from a series of turbulent jets. For evaluation on the synthetic data, results are compared to other optical flow methods (using scalar images) found within the literature and correlation-based PIV using synthetic tracer particle images generated from the same DNS. For scalar-based velocimetry, there are two primary sources of error that do not occur for correlation-based PIV or when wOFV is applied to tracer particle-based images: (i) low occurrences (spatially isolated) of high-magnitude errors due to motion ambiguity resulting from poor local contrast in the scalar images and (ii) out-of-plane scalar flux. Overall, the application of the wOFV method to scalar images is less accurate than tracer particle-based approaches (correlation-based PIV or wOFV), but still yields reasonably accurate velocity estimations. The current scalar-based wOFV method compares very favorably to previous scalar image-based optical flow methods and may find utility in cases where it is undesirable or impossible to seed the flow with tracer particles. A parametric study is performed to determine experimental design parameters, including optimal conditions for spatial resolution (i.e., image contrast) and maximum inter-frame displacement. Data from a three-dimensional DNS is used to directly assess errors caused by scalar transport in the direction perpendicular to the imaging plane, which could be present in laser-based planar measurements. Finally, the scalar-based wOFV method is applied to experimental image sequences of passive scalar fields from a series of turbulent jets to serve as proof-of-concept. Initial results show that wOFV may be suitable for velocimetry on experimental scalar images if there is a primary flow direction (i.e., sufficiently two-dimensional flows).