This project supports construction of a novel computational imaging system for efficient and robust acquisition and reconstruction of multi-scale flows. Since transparent flows do not have their own images but borrow appearance from nearby objects, the solution developed here directly acquires light paths off/through the flow wavefront/volume and then infers the physical properties of the flow. The acquisition system is portable and non-intrusive, to support on-site acquisition of various types of flows. Time-dependent 3D transparent flows such as fluid wavefronts, gas-liquid interfacial flows, and turbulent flows are ubiquitous in chemical, biological and environmental engineering. Accurate reconstruction of such flows under experimental setups have a broader impact on these areas and their applications, e.g., studying coal combustors and predicting natural phenomena such as warm rain and hurricane.

The new 3D flow acquisition system leverages emerging light field (LF) cameras and displays. An LF collects all rays emitting from a 3D scene. The research team explores two unique properties of LFs, ray sampling and multi-view imaging. A reconfigurable LF camera-display system is first developed to directly capture how the flow changes the light paths. Tailored computer vision algorithms are then developed to infer physical properties of the flow (density, velocity, vorticity, etc). Compared with state-of-the-art solutions, the new acquisition system is portable and can be re-configured to acquire various flow types ranging from completely specular wavefronts to transparent gas volumes and to turbulent flows laden with solid particles. For validation, direct (with all scales resolved) simulations of turbulent channel flows are conducted, without and with solid particles, against the experimental measurements using a mesoscopic approach known as the lattice Boltzmann (LB) method.

Project Start
Project End
Budget Start
2015-06-15
Budget End
2017-05-31
Support Year
Fiscal Year
2015
Total Cost
$174,445
Indirect Cost
Name
University of Delaware
Department
Type
DUNS #
City
Newark
State
DE
Country
United States
Zip Code
19716