Genetic IMM_NN based tracking of multiple point targets in infrared image sequence

Abstract

Tracking of maneuvering and non-maneuvering targets simultaneously is a challenging task for multiple target tracking (MTT) system. Interacting multiple model (IMM) filtering has been used for tracking multiple targets successfully. IMM needs to evaluate model probability using an observation assigned to the track. We propose a tracking algorithm based on IMM which exploits the genetic algorithm for data association. Genetic algorithm performs nearest neighbor (NN) based data assignment. A mixture probability density function (pdf) for the likelihood of the observation is used for data assignment.