Analysis of MHT and GBT Approaches to Disparate-Sensor Fusion

Abstract

Multi-sensor multi-target tracking requires the solution to a challenging data association problem. The problem simplifies when a portion of the target state vector and the corresponding sensor data satisfy a particular Markovian assumption. This leads to quantifiable benefits in performance vs. complexity of the tracking solution. This paper summarizes recently-obtained technical advances in graph-based tracking and applies this to a benchmark study with respect to an advanced track-oriented multiple-hypothesis tracking solution.