High-resolution three-dimensional flow measurement through dual-frame light field particle tracking velocimetry

Single-camera light field particle image velocimetry (LF-PIV) shows potential for three-dimensional (3D) flow measurements in scenarios with limited optical access but faces challenges of low spatial resolution. To address this issue, we propose a dual-frame light field particle tracking velocimetry (LF-PTV) method that enhances spatial resolution in volumetric velocimetry. This approach combines line-of-sight estimation with a customized deep neural network to reconstruct particle volumes while suppressing elongation artifacts. A gradient-fitting localization technique is employed to pinpoint particle centers, and a motion predictor coupled with a topology-feature matching method facilitates accurate trajectory tracking between successive frames. The performance of the dual-frame LF-PTV method is systematically evaluated through numerical simulations of Gaussian vortex flows and experimental measurements of wake flow behind a circular cylinder. Comparative analyses are conducted to benchmark the proposed method against other PTV and conventional LF-PIV techniques. Results indicate that the deep neural network effectively refines coarse line-of-sight reconstructions, significantly reducing particle elongation. The deep neural network reconstruction using a single light field camera (LFC) achieves approximate accuracy with the traditional Simultaneous Multiplicative Algebraic Reconstruction Technique using dual LFCs. The gradient-fitting algorithm can achieve superior particle localization, especially in high-density seeding, by reducing outliers and enhancing coverage. Furthermore, the integration of motion prediction with the topology-feature matching approach minimizes tracking errors, yielding superior accuracy and spatial resolution in synthetic vortex flow reconstructions. Experimental results further confirm the method's capability to resolve finer wake flow structures, correcting LF-PIV inaccuracies and delivering a three times enhancement in spatial resolution.