Three-dimensional point pattern tracking using a completely determined Hopfield network independent of the input data

Summary form only given, as follows. Three-dimensional point pattern tracking is viewed as an optimization problem where a global cost function, integrating multiple constraints on the feature points to be matched, is minimized. Minimization is efficiently performed by a Hopfield net independent of the input data, and the optimum positive constants of the cost function are determined in advance based on the stability of the network at equilibrium and its speed of convergence. Tested with computer generated data, the simulated network shows good performance, especially in the presence of noise, and compares favorably with other matching procedures.<>