Multi-Frame Track-Before-Detect Algorithm for Maneuvering Target Tracking

Abstract

Multi-frame track-before-detect (MF-TBD) is a model-based batch processing method. Assuming a particular model for the evolution of target states (e.g. a constant velocity model) within a batch processing time, MF-TBD integrates the target energy by taking advantage of the space-time correlations between a number of consecutive frames. Its performance is known to be heavily dependent on the accuracy of the motion model, and to substantially degrade when target maneuvers occur and motions do not follow the presumed model. We make two contributions towards addressing this problem. Firstly, we analyze and summarize the direct strategies to incorporate the effect of target maneuvers into MF-TBD. Our analysis shows that although these strategies are straightforward to implement, they either suffer from considerable performance loss or are computationally expensive. Motivated by the analysis, we propose a general measurement-directed (MD) strategy to address target maneuvers. It carries out an on-line study of target dynamics from the observations, and is capable of achieving both low computational complexity and high adaptability to different target maneuvers. Secondly, as the proposed MD strategy is a general framework without any particular model assumptions, we further derive its detailed implementation equations for linear motion and measurement models. Simulation results for various tracking scenarios are presented to demonstrate the effectiveness of the proposed MD strategy.