Tracking algorithm using N-back scan MHT under dense environments

MHT(multiple hypothesis tracking) is well-known tracking algorithm for its association performance in dense environments. In the MHT, hypotheses are reduced by calculating probability and upper hypotheses are selected in a usual way to reduce computational burden. In the field of visual tracking, the N-back is applied as hypothesis reduction method and its efficiency is verified. The method maintains all hypotheses in recent N scans and selects the best hypothesis in past over N-scan, for hypotheses reduction. We apply the method to radar tracking and evaluate its performance, through Monte Carlo simulation of track maintenance for a low observable target. From the result of simulation, we have confirmed that N-back scan MHT shows better performance than conventional MHT. For example, the former has shown 30 point improvement with respect to tracking success rate for tracking a maneuvering target than latter.