一种高效的分布式多传感器多目标跟踪算法

现有的多传感器多目标跟踪算法大都基于马尔科夫-贝叶斯模型,需要诸如目标运动、杂波、传感器检测概率等先验信息,但是在恶劣的环境中,这些先验信息不准确并导致目标跟踪精度下降。为了解决该情况下的多目标跟踪问题,我们提出了一个高效的分布式多目标跟踪算法,该算法通过泛洪(Flooding)共识算法在分布式网络的传感器之间迭代的传输、共享各自的量测集信息,并通过改进的密度峰值聚类(Improved Density Peaks Clustering, IDPC)算法对量测集聚类,聚类得到的簇的个数即目标的个数,簇的中心即目标的位置。我们将IDPC算法与前沿的分布式概率密度假设(probability density hypothesis, PHD)滤波器在三个场景中进行对比,实验结果证明了IDPC算法的有效性和可靠性。 

Research on Distributed Multi-Sensor Multi-Target Tracking Algorithm

The existing multi-sensor multi-target tracking algorithms are mostly based on Markov-Bayes model, which requires prior information such as target motion, clutter, and sensor detection probability, but in harsh environments, these prior information are not accurate and lead to a decrease in target tracking accuracy. To solve the MTT problem in such situation, we propose an efficient distributed multi-target tracking algorithm, which uses a flooding consensus algorithm to iteratively transmit and share the measurement set information between sensors in the distributed network, and cluster the measurement set through an Improved density peaks clustering (IDPC) algorithm. The number of clusters obtained by clustering is the number of targets, and the center of the clusters is target’s position. We compare the IDPC algorithm with cutting edge distributed probability density hypothesis (PHD) filters in three scenarios, the experimental results prove the effectiveness and reliability of the IDPC algorithm.