Reducing computational complexity of gating procedures using sorting algorithms

Gating is an important part in many data association algorithms, specifically tracking algorithms. Its purpose is the preselection of suitable measurements or observations respectively in order to avoid unlikely measurement-to-track associations. As such, it can be essential for a low computational load. A very common gating method is individual ellipsoidal gating incorporating the Mahalanobis distance. On the one hand it is a reliable gating method used in many different tracking techniques, e.g., the global nearest neighbor method or the Multi-Hypothesis Tracking approach, but on the other hand it can be also very time consuming due to the inversions of covariance matrices related to all the measurement-to-track pairs. In this paper a Gating-by-Sorting approach is proposed to accelerate the gating procedure significantly. It uses sorting algorithms to carry out coarse gating followed by a second ellipsoidal gating step. The gating approach is incorporated into a multi-hypothesis tracker and applied to a real radar dataset. Results are obtained by measuring the computational load within tracking processes. They show that the proposed gating algorithm outperforms classical ellipsoidal gating by orders of magnitude.