基于随机有限集的可分辨群目标跟踪算法研究

发布时间：2018-07-04 09:24

本文选题：可分辨群目标 + 图理论　；参考：《杭州电子科技大学》2017年硕士论文

【摘要】：群目标由多个相互协作的目标组成,并且保持着一定结构进行运动。随着传感器技术的不断发展,原来不可分辨的群目标呈现出可分辨的属性,群目标跟踪算法的研究也越来越重要。跟踪估计主要包括获得目标的量测,并对量测进行滤波和估计。因此,需要对群目标进行动态建模,并采用合适的算法对群目标进行跟踪估计。现在对群目标的动态建模主要是将群目标中的各成员的状态进行简单的结合,但是该方法不能充分体现出群目标各成员之间的依赖关系。目前的群目标跟踪研究主要集中在不可分辨群目标或者部分可分辨群目标上,如果将不可分辨群目标的跟踪方法应用于可分辨群目标,会导致估计过于粗糙从而丢失许多结构信息。为了解决上述问题,本文使用图理论结合标签随机有限集(L-RFS)理论,并且引进目标的可分辨系数,给出在可分辨情况下群目标新的动态建模方法和跟踪估计方法,主要工作包括:(1)可分辨群目标的动态建模。根据群结构与图结构的相似性,利用描述有向图的方法对群进行描述,即使用邻接矩阵来描述群。再通过邻接矩阵判断群中各成员的依赖关系。由于头结点起到领导作用,因此可以对头结点进行独自建模。再建立其它依赖于头结点的子结点的动态模型,并依次建立其它结点的动态模型。(2)群目标可分辨系数。由于雷达的分辨率会随着探测距离的增大而变低,使获得的量测误差变大。假设该误差服从正态分布,而正态分布的均值到边缘的距离大概为标准差?的3倍,因此,本文令群目标的可分辨距离i,jr为3 3i j?(10)?,其中i和j表示目标i和目标j。当群目标中各对成员的距离大于i,jr时,认为群目标是可分辨的。(3)可分辨群目标的跟踪算法的研究。第一步,由于起始阶段群目标之间的协作关系未知,因此假设群目标之间是独立的并采用广义标签多伯努利(GLMB)滤波算法获得各目标的状态估计和轨迹估计以及目标的个数估计。第二步,在获得群目标中各成员的状态估计基础上,通过计算每时刻的偏差矩阵估计获得邻接矩阵估计,并通过邻接矩阵得到群的结构关系。再利用图论中连通图的概念获得各时刻子群的个数估计。
[Abstract]:Group targets are composed of multiple cooperative targets and move in a certain structure. With the continuous development of sensor technology, the original indistinguishable group targets show discernible attributes, and the research of swarm target tracking algorithm is becoming more and more important. Tracking estimation mainly includes obtaining the measurement of the target and filtering and estimating the measurement. Therefore, it is necessary to dynamically model the group target and use the appropriate algorithm to track and estimate the group target. At present, the dynamic modeling of group objects is mainly based on the simple combination of the states of each member of the group object, but this method can not fully reflect the dependency relationship among the members of the group object. The current research on group target tracking is mainly focused on indiscernible group targets or partially discernible group targets. If the tracking method of indiscernible group targets is applied to discernible group targets, This can result in rough estimates and the loss of a lot of structural information. In order to solve the above problems, this paper uses graph theory combined with label random finite set (L-RFS) theory, and introduces the discernibility coefficient of targets, and gives a new dynamic modeling method and tracking estimation method for group targets under discernible conditions. The main work includes: (1) dynamic modeling of discernible group targets. According to the similarity between group structure and graph structure, the method of describing directed graph is used to describe the group, that is, the adjacent matrix is used to describe the group. Then the dependency of each member in the group is judged by the adjacency matrix. Because the head node plays a leading role, we can model the head node alone. Then the dynamic models of other child nodes dependent on the head node are established, and the dynamic models of the other nodes are established in turn. (2) the discernibility coefficient of the group targets. Because the resolution of radar decreases with the increase of detection range, the measurement error becomes larger. Suppose the error is from normal distribution, and the distance from the mean to the edge of the normal distribution is approximately standard deviation? Therefore, in this paper, the discernible distance ig jr of the group target is 3 3i jr (10), where I and j denote the target I and the target j. When the distance of each pair of members in a group target is greater than that of ijr, it is considered that the group target is distinguishable. (3) the research of tracking algorithm for discernible group target In the first step, because of the unknown cooperative relationship between group targets in the initial stage, it is assumed that the group targets are independent and the generalized label multi-Bernoulli (GLMB) filtering algorithm is used to obtain the state estimation and trajectory estimation of each target and the estimation of the number of targets. In the second step, on the basis of the state estimation of each member of the group target, the adjacent matrix estimation is obtained by calculating the deviation matrix estimation at every moment, and the structure relation of the group is obtained by the adjacent matrix. Then, by using the concept of connected graph in graph theory, the number of subgroups at each time is estimated.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN713

【参考文献】