基于最大熵直觉模糊核聚类的目标跟踪算法

Target tracking based on maximum entropy intuitionistic fuzzy Kernel clustering

为了提高杂波环境下目标跟踪的正确关联率和实时性,本文提出一种基于最大熵直觉模糊核聚类的目标融合跟踪算法。先通过密度函数法确定初始聚类中心,再通过加入核函数和放松对隶属度的限制,并且通过样本加权给离群点和样本点不同的权值,从而可以减少离群点和噪声点的干扰,最后通过直觉指数引入直觉模糊集,得到改进后隶属度矩阵,以隶属度矩阵作为关联概率进行目标与观测的关联,并用卡尔曼滤波进行目标模型的更新,提高目标跟踪的正确关联率和实时性。实验表明,本文算法相对传统的模糊C均值聚类算法可以提高目标正确关联率3%左右,并且在算法耗时方面平均减少了0.1 s,相对于最大熵模糊C均值聚类算法可以提高目标正确关联率1.9%左右,算法耗时平均减少0.4 s,表明本文算法在提高目标跟踪正确关联率并增加算法实时性拥有更好的效果。

In order to improve the correct correlation rate and real-time performance of target tracking in clutter environment,this paper proposes a target fusion tracking algorithm based on maximum entropy intuitionistic fuzzy kernel clustering.Firstly,the initial clustering center is determined by density function method;secondly,the kernel function is added and the restriction on membership degree is relaxed;finally,the interference of outliers and noise points can be reduced by weighting samples;the intuitionistic fuzzy set is introduced by intuitionistic index to obtain the improved membership degree matrix;finally,the membership degree matrix is used as the correlation probability to associate the target with the observation;finally,the Kalman filter is used to update the target model to improve the correct correlation rate and real-time performance of target tracking.Experiments show that the proposed algorithm can improve the correct correlation rate of targets by about 3%and reduce the average time consumption by 0.1 s compared with the traditional fuzzy C-means clustering algorithm.Compared with the maximum entropy fuzzy C-means clustering algorithm,it can improve the correct correlation rate of targets by about 1.9%and reduce the average time consumption by 0.4 s,which shows that the proposed algorithm has better effect in improving the correct correlation rate of target tracking and increasingthe real-time performance ofthe algorithm.