SNR limits on Kalman filter detect-then-track

Abstract

In the common parlance of target tracking literature, conventional tracking refers to the sequential application of a single-frame detector to sensor data and a point-measurement association and estimation (tracking) algorithm. In contrast, Track-Before-Detect (TkBD) refers to the processing paradigm where detection and estimation are performed jointly. The TkBD literature reports significant sensitivity improvements over conventional tracking and tends to attribute this to losses at the single-frame detector. This article shows that this is not the case but rather the poor performance of conventional tracking at low signal power is more attributable to the Kalman Filter.