Issues in single target tracking with state dependent detectionprobability

摘要

Target tracking algorithms have traditionally assumed that thetarget detection probability is uniform across the validation gate. Thispaper examines one approach for single target tracking when the targetdetection probability is non-uniform (state-dependent) across the gate.By using the expected value for the target measurement a trackerformulation similar to probabilistic data association (PDA) is derivedwith the summations in PDA replaced by integrals over the a posterioriinnovation probability density function in the state-dependentformulation. The reduction of the new tracking update equations to theconventional formulation is shown along with a discussion of the effectof restricting target innovations of interest to the validation gate.The behaviour of the new tracker for a partially obscured validationgate is also demonstrated