【作者】 方亮；

【导师】 杨宏文；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2009， 硕士

【摘要】 航迹相关是多传感器融合跟踪系统的关键技术之一,系统偏差的存在使航迹相关问题变得更为复杂。传统的处理方式是先估计出系统误差的值,然后再进行航迹相关。论文重点研究在不估计出系统误差具体数值的情况下直接进行航迹相关的方法,主要内容如下:1、研究了基于历史信息的航迹相关方法。由于系统误差等因素的影响,不同传感器观测到的同一个目标的航迹之间存在一定的位置偏差。这一位置偏差的取值范围可作为航迹相关的初步判定依据。经过一段时间的跟踪后,可以建立航迹间的位置偏差序列。经分析,位置偏差序列的样本方差存在理论最大值,基于样本方差理论最大值和实际样本方差的比值,提出了基于历史信息的航迹相关方法。2、将航迹看成是空间中的点模式,利用图谱理论和灰色理论,研究了基于航迹空间分布信息的航迹相关方法。首先研究了归一化Laplace矩阵作为表示矩阵的图谱相关算法;然后研究了一种利用局部传感器观测到的航迹数据构建灰色自关联矩阵的灰色相关方法,该方法在有效利用航迹空间分布信息的基础上,可以将不同测度的属性信息融入到航迹相关过程中来,具有一定的实用价值。3、论文最后研究了将航迹历史信息和空间分布信息结合起来的航迹相关算法,仿真实验表明,相比利用单一信息的相关方法,该方法性能更好。更多还原

【Abstract】 Track correlation is one of the key technologies in multisensor-multitarget tracking and fusion system. However, the existence of system bias made the track correlation problem more difficult to solve. The common technologies deal this problem with classical track correlation method after estimating the system bias. This thesis concentrates on the novel track correlation method without estimating the system bias. The main contribution can be summarized as follows:1. A track correlation method based on history information is researched in chapter2 Tracks of one target measured by different sensors in different location have position errors which is caused by coordinate conversion with system bias. By deducing the bounds of position errors, the coarse correlation method is proposed. It is demonstrated in the thesis that there exists a theoretical maximum variance of position errors sequence which is obtained by time accumulation. Depending on the empiric to theoretical variance ratio, the precision correlation method for track correlation is presented.2. Tracks can be regarded as the point pattern in space at different times, so the graph theory and the gray theory can be introduced for track correlation based on the distributed space information. The generalized Laplace matrix is researched in chapter 3 for track correlation, and a correlation method which constructs the gray correlation matrix by local sensor measures is proposed. This novel method can use the distributed space information sufficiently and fusion the attribute information with different measurement conveniently.3. The thesis presents a track correlation algorithm which is based on the combination of both the information of history and space. The simulation results demonstrate that the correlation performance is better than using just information of history or space.更多还原