基于模糊聚类的多目标跟踪算法  

Multiple Targets Tracking Algorithm based on Fuzzy Clustering

作　　者：陈出新[1] 周德云[1] 

机构地区：[1]西北工业大学电子信息学院,西安710072

出　　处：《火力与指挥控制》2010年第2期89-91,共3页Fire Control & Command Control

摘　　要：提出了一种新的基于模糊聚类的多目标跟踪算法,该算法通过一种改进的模糊聚类算法,首先得到可能的目标数和测量点迹与目标预测位置之间的隶属度,然后结合Kalman滤波将隶属度作为权值系数对预测新息向量进行加权,来实现目标状态估计的更新。仿真结果表明,传统数据融合多目标跟踪算法,一般需要假定目标数并且在多目标密集时易产生关联错误而导致跟踪发散,新算法通过模糊聚类客观有效地确定了目标数并且通过加权过程保证了对多目标密集时的高精度。A novel multiple targets tracking algorithm based on fuzzy clustering is proposed. This algorithm is to obtain the number of targets and degrees of membership between the measurements and the predicted locations of the targets through an improved fuzzy clustering algorithm. It uses the degrees of membership between the measurements and the predicted locations of the targets which are obtained through fuzzy clustering as weight numbers of innovations to realize the update of targets＇ states. The simulation results demonstrate that the conventional algorithm needs to suppose the number of targets and easily to make a wrong association when targets are near to each other which may lead to the emanative tracking. However, the novel algorithm is to determine the number of targets objectively and to guarantee high quality of tracking through the weighted sum of innovations.

关 键 词：模糊聚类 多目标跟踪 KALMAN滤波 数据融合 

分 类 号：TN953[电子电信—信号与信息处理]