Elastic tracking and neural network algorithms for complex pattern recognition

Abstract

A new elastic tracking (ET) algorithm is proposed for finding tracks in very high multiplicity and noise environments. It is based on a dynamical reinterpretation and generalization of the Radon transform and is related to elastic net algorithms for geometrical optimization. ET performs an adaptive nonlinear fit to noisy data with a variable number of tracks and is more efficient numerically than the traditional Radon or Hough transform method, because it avoids binning of phase space and the costly search for valid minima. Spurious local minima are avoided in ET by introducing an iteration time-dependent effective potential. The method is shown to be very robust to noise and measurement error and extends tracking capabilities to much higher track densities than possible via local road finding or even the novel Denby-Peterson (DP) neural network tracking algorithms. A possible neural network implementation of ET is also discussed.