A hybrid 3D particle matching algorithm based on ant colony optimization

Autor: Mingyuan Nie, Jinjun Wang, Chujiang Cai, Chong Pan
Rok vydání: 2021
Předmět:
Zdroj: Experiments in Fluids. 62
ISSN: 1432-1114
0723-4864
0957-0233
DOI: 10.1007/s00348-021-03160-4
Popis: Particle Tracking Velocimetry (PTV) is a popular optical method to measure the velocity field of complex flow at high spatial resolution. One of the practical limitations of this technique is that in general, the accuracy of matching individual particles between a pair of images is limited by the particle image displacement. To deal with this limitation, the present work proposed a hybrid ant colony optimization (ACO) algorithm for particle matching in three-dimensional (3D) scenarios . It can be regarded as an update of the conventional ACO PTV algorithm (Takagi J Vis 27:89–90. https://doi.org/10.3154/jvs.27.Supplement2_89 , 2007, Ohmi et al. Exp Fluids 48(4):589–605. https://doi.org/10.1007/s00348-009-0815-2 , 2010). The key concept is to seek a global solution of the minimization of a displacement-pattern function (DPF) via improved ant colony optimization (ACO). The object function, i.e., DPF, hybrids the measure of particle image displacement and the measure of pattern similarity as the particle matching criterion, the latter of which is constructed as the similarity level of Voronoi polygons (VPs) of paired particles. Performance evaluation was based on both the standard particle image database of Visualization Society of Japan (Okamoto et al. Meas Sci Technol 11(6):685–691. https://doi.org/10.1088/0957-0233/11/6/311 , 2000) and the laboratory-made synthetic flow. It was shown that this hybrid ACO algorithm has higher matching accuracy than those of exiting ACO methods based on either minimum displacement function or relaxation function. Its credibility in dealing with the scenarios of large relative particle displacement, i.e., the cases where particle image displacement is comparable to or even larger than the mean spacing of neighboring particles was also empirically demonstrated. Other features including fast convergence speed and regular pattern of outliers were also seen. All these make this algorithm a suitable candidate for 3D particle matching in PTV.
Databáze: OpenAIRE
Externí odkaz: