基于观测数据聚类划分的扩展目标跟踪算法

发布时间：2018-06-11 11:37

本文选题：多目标跟踪 + 扩展目标跟踪　；参考：《数据采集与处理》2017年01期

【摘要】：针对高斯混合假设概率密度滤波扩展目标跟踪算法中的观测集合划分问题,本文分析了距离门限划分方法存在的问题,提出了一种利用聚类方法进行观测数据集合划分的方法。该方法利用极大似然方法获得目标数的估计值,以此作为聚类数利用K-means聚类对观测数据集合进行聚类划分,并利用椭球门限对观测数据进行处理以降低杂波观测对聚类划分的影响。实验结果表明,该方法能够获得正确的划分观测集合,计算复杂度较距离门限划分方法有较大降低,同时在多目标跟踪效果方面优于观测数据距离门限划分的假设概率密度滤波扩展目标跟踪算法。
[Abstract]:Aiming at the problem of the partition of observation set in the extended target tracking algorithm of Gao Si hybrid hypothesis probability density filter, this paper analyzes the problems of the distance threshold partition method, and puts forward a method to partition the observation data set by using the clustering method. In this method, the maximum likelihood method is used to obtain the estimated value of the target number, which is used as the clustering number to classify the observation data set by K-means clustering. The ellipsoidal threshold is used to process the observed data to reduce the influence of clutter observation on clustering. The experimental results show that this method can obtain the correct partition observation set, and the computational complexity is much lower than that of the distance threshold partition method. At the same time, the extended target tracking algorithm based on probability density filter is better than the threshold partition of observation data in the aspect of multi-target tracking.
【作者单位】：中国民航大学天津市智能信号与图像处理重点实验室;天津大学电子信息工程学院;
【基金】：国家自然科学基金(61471365,61231017,61571442)资助项目中国民航大学中央高校基金(3122015D003)资助项目
【分类号】：TN713

【相似文献】