利用可能模糊聚类的鲁棒目标跟踪方法  

Robust target tracking method based on possibility fuzzy clustering

作　　者：蔡秀梅 王妍 吴成茂 卞静伟 CAI Xiumei;WANG Yan;WU Chengmao;BIAN Jingwei(School of Automation,Xi'an University of Posts and Telecommunications,Xi'an 710121,China;School of Electronic Engineering,Xi'an University of Posts and Telecommunications,Xi'an 710121,China)

机构地区：[1]西安邮电大学自动化学院,陕西西安710121 [2]西安邮电大学电子工程学院,陕西西安710121

出　　处：《西安邮电大学学报》2020年第4期53-59,共7页Journal of Xi’an University of Posts and Telecommunications

基　　金：陕西省重点研究开发项目(2019GY-107)。

摘　　要：为了提高噪声情况下目标跟踪的准确度,提出一种基于直觉可能性模糊C均值聚类算法。首先,引入可能性模糊C聚类算法对含有噪声的目标进行聚类,得到分别以目标和观测为聚类中心的隶属度矩阵。其次,结合直觉模糊集得到直觉可能性模糊隶属度矩阵,对得到的隶属度矩阵根据最大隶属度原则求出目标和观测的正确关联对。最后,应用Kalman滤波更新目标轨迹,实现噪声情况下的目标跟踪。实验结果表明,在噪声环境下,改进算法目标跟踪的准确度和鲁棒性较好。In order to improve the robustness of target tracking under noise,a fuzzy clustering algorithm based on intuitionistic possibility is proposed.Firstly,the possibility fuzzy clustering algorithm is introduced to cluster noisy targets to obtain membership matrices with targets and observations as clustering centers respectively.Then,the intuitionistic possibility fuzzy membership matrices are obtained by combining the intuitionistic fuzzy sets.The correct correlation pairs between targets and observations can be obtained by applying the principle of maximum membership to the obtained membership matrices.Finally,the Kalman filter is applied to update the trajectories in order to realize target tracking under the condition of noise.Experimental results show that the improved algorithm can effectively improve the accuracy and robustness of target tracking under noise conditions.

关 键 词：直觉模糊聚类 可能性模糊C均值聚类 卡尔曼滤波 目标跟踪 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]