A Real-Time Filtering Approach for Plot-Sequences Outputted by Multi - Frame Track - Before- Detection

摘要

This paper addresses the real-time filtering problem respect to the multi-frame track-before-detection (MF - TBD). Instead of making detections on each single frame, MF - TBD jointly processes multiple raw data frames at each time, which can remain and utilize more target information. Thus, different from the conventional detection procedure which provides single point plot to a tracker, output of MF - TBD is plot-sequence. However, for batch processing MF - TBD, there are repetitive parts (named plot-sets) under a number of consecutive batches in time series while the correlation among each plot in the plot-set is complex and unpractical to computation. In this paper, by avoiding the computation of correlation, a novel filtering approach is developed for the plot-sequences outputted by batch processing MF - TBD. Last, simulation results show that the proposed algorithm can correctly estimate target trajectories and significantly enhance the tracking accuracy compared with the MF - TBD without our filtering algorithm.