A computational network for global optimization of particle tracks in stereo image velocimetry

A computational network is shown to determine globally optimal tracks in stereo image velocimetry. Data extracted from two-dimensional particle images are mapped onto a highly interconnected network of processing elements. The data, network constraints, and flow dynamics provides the information required to track seed particles. The combinatorial complexity of particle tracking is avoided by equations of motion which efficiently guide the network to a stable solution. Particle overlap is overcome by mapping the results of probability based overlap decomposition onto the network. The algorithm is self-starting and self-terminating. Results of experiments are presented to demonstrate the efficacy of the method.