SOM BASED PARTICLE MATCHING FOR VOLUMETRIC PARTICLE TRACKING VELOCIMETRY

摘要

Novel 3D image analysis and particle matching techniquesrnfor the use in the volumetric particle tracking velocimetryrnhave been developed and tested by using synthetic imagesrnand experimental images of unsteady 3D flows. A tomog-rnraphy based particle reconstruction scheme along with thernsubsequent process of individual particle detection and val-rnidation was used. The detected particles in the two timerndifferential samples are matched by using Self OrganisingrnMap (SOM) neural network scheme. SOM neural networkrntracking algorithm is highly adaptive to time differentialrntracking even with loss-of-pair particles. The particle loca-rntion and velocity results of the present new approach turnedrnout accurate, reliable and robust in comparison to the con-rnventional 3D PTV approaches.