Stereoscopic Correspondence of Particles for 3-Dimensional Particle Tracking Velocimetry by using Genetic Algorithm

Abstract

Abstract:The genetic algorithm (GA) based stereo particle-pairing algorithm has been developed and applied to the spatial particle-pairing problem of the stereoscopic three-dimensional (3-D) PTV system. In this 3-D PTV system, particles viewed by two (or more than two) stereoscopic cameras with a parallax have to be correctly paired at every synchronized time step. This is important because the 3-D coordinates of individual particles cannot be computed without the knowledge of the correct stereo correspondence of the particles. In the present study, the GA algorithm is applied to the epipolar line proximity analysis for establishing correspondence of particles pairs between two co-instantaneous stereoscopic particles images, in order to compute the 3-D coordinates of every individual particle. The results are tested with various standard images and it's found that the new strategy using GA works better than conventional particle pairing methods of 3-D particle tracking velocimetry for steoroscopic PTV.

A spatial-temporal algorithm for three-dimensional particle tracking velocimetry using two-view systems

“…The proposed algorithm is compared with a five-frame fully automated (FA) 3D PTV algorithm [6] and other four AI algorithms including GA [15], ACO [19], CNN [21] and SNN [20], which are all performed with two-view imaging systems using ground-truth synthetic particle positions. The synthetic data of the GA [15], ACO [19] and CNN [21] methods are generated from the DNS results of a 3D impinging jet in a square cavity, while SNN [20] and FA method [6] adopt two 3D vortical flows, respectively. The fractions of correctly detected particles and vectors using different methods are presented in figures 12(a) and (b).…”

“…With the smoothed trajectory, predictions are made using the velocity and acceleration information obtained from previous frames using equations (15) and (16). At this stage, the established trajectories are highly reliable.…”

“…In a two-view system, the epipolar constraint can only limit the search area around a line rather than the intersecting point. Thus, it can only handle relatively low particle densities, normally less than 0.005 ppp [6,15]. In addition to the direct imaging system, Wang and Zhang [16] have established a two-view shadowgraph system using the pinhole camera model to study bubble bursting.…”

“…The ant colony optimization (ACO) method has been employed to find stereo particle pairs by minimizing the sum of normal distances between particle centers and epipolar lines of two views, and reasonable results were reported around 0.005 ppp [19]. Panday introduced a genetic algorithm (GA) to solve such an optimization problem in stereo matching, which performs better than the ACO method when the particle image density is larger than 0.015 ppp [15]. Gim et al [20] have applied a shallow neural network (SNN) model to recover particle 3D positions with 99.9% reconstruction accuracy at 0.004 ppp.…”

Volumetric particle tracking velocimetry with improved algorithms using a two-view shadowgraph system

“…These AI-based tracking methods adopt a reconstruction-then-tracking strategy, which reconstruct the 3D particle positions first and then track the 3D trajectories in space. Ant colony optimization method [22], genetic algorithm [23], cellular neural network [24], shallow neural network [25], have been applied to two-view systems to recover particle positions and velocities optimally. Most of these methods have shown a high correctness over 90% when the particle seeding density is no more than 0.005 ppp, above which it drops to a lower percent.…”