Low Complexity δ‐GLMB Maneuvering Multi‐Target Detection and Tracking for High‐Dimensional Composite Hypothesis

ABSTRACT The use of single‐frame data for state updates in traditional tracking algorithms is prone to significant errors when tracking multiple maneuvering targets in complex environments. Although Multiple Hypothesis Tracking (MHT) algorithm takes into account multi‐frame information, the computation amount will heighten sharply when the number of targets raises and the complexity of environmental factors gets more noticeable. To cope with this problem, a low complexity δ‐Generalised Labelled Multi‐Bernoulli (δ‐GLMB) maneuvering multi‐target tracking method for high‐dimensional composite hypothesis is proposed in this paper. Owing to the use of multi‐frame information, the δ‐GLMB filter can effectively put an end to the problem of undesirable tracks in complex scenes. More importantly, the calculation complexity of MHT algorithm can be lessened by forming multiple hypotheses through target assemblies. Simultaneously, the recommended algorithm is extended to non‐ideal detection conditions with high clutter density and low detection probability, which is advantageous for realizing the multi‐target detection and tracking combined with Track‐Before‐Detect (TBD) algorithm. As suggested by simulation results, the low complexity δ‐GLMB maneuvering multi‐target tracking method for high‐dimensional composite hypothesis can effectively track targets in complex multi‐target scenes. Aside from that, the combination of TBD makes the algorithm have better multi‐target detection performance and more stable multi‐target tracking effect in non‐ideal scene with high clutter density and low detection probability.