An Efficient Recursive Multiframe Track-Before-Detect Algorithm

Abstract

Multiframe track-before-detect (MF-TBD) usually uses sliding-window-based batch processing, where a number $N$ of the latest data frames are jointly processed at each measurement time. The sliding window mechanism compromises the operating efficiency of MF-TBD by increasing both computational costs and memory requirements, thus, heavily restricting its application in practical problems. In this paper, an improved recursive implementation for MF-TBD is proposed. Unlike the sliding-window-based implementation, the proposed method calculates the merit function, a measure of the possibility that a state is target originated, of the current batch based on an approximated recursive relationship between the merit functions of consecutive batches. As a result, instead having to process the whole batch, at any given time only the latest frame needs to be processed. The recursive relationship is first derived for any arbitrary merit function, and then explored further with several typical merit functions that are used in MF-TBD. Both the theoretical analysis and simulation results demonstrate that the proposed method can achieve almost $N$ times reduction in computational complexity and memory requirements with negligible performance loss.