关键词: 概率多假设跟踪, 纯方位, 多目标跟踪, 无味卡尔曼平滑算法

Abstract:

The weakness of the basic Probabilistic Multiple Hypothesis Tracking （PMHT） is in its sensitivity on the targets’ initialization. Once the estimated initialization is different from the real initialization, the tracking performance of the PMHT is degraded seriously. Based on the principle of PMHT and the deterministic annealing approach, combining the Unscented Kalman Smoother（UKS） algorithm, it proposes a new multi-target tracking algorithm: deterministic annealing PMHT. It compares and analyzes the estimation accuracy and calculation cost of the deterministic annealing PMHT and PMHT in simulation experiments for underwater bearing-only cross-moving targets and closely spaced targets under the dense clutter environment. The experimental results demonstrate that PMHT algorithm’s tracking performance degrades severely when setting a relatively bad initialization value, while the correct association of deterministic annealing PMHT is of high probability. The deterministic annealing PMHT has effective detection performance and noise tolerance, and it has little computational load at the same time, and confirms the effectiveness of the deterministic annealing PMHT to the bearing-only multiple target tracking in dense clutter.

Key words: probabilistic multiple hypothesis tracking, bearing-only, multi-target tracking, unscented Kalman smoother