摘要
将特征参数测量值引入概率数据关联(PDA)及滤波算法中,提出一种计算互联概率的象素级融合算法。本算法采用的是双滤波器结构,二者使用共同的互联概率。并通过仿真验证了本算法对某些模型性能良好。
In multiple targets tracking, attribute parameters can be used to improve the performance of data association. B.J.Slocumb used augmented state vector to incorporate attribute parameters in estimation process. But the computation overhead increases very fast with the number of attribute parameters incorporated. We present a new method to solve this problem. Both state parameters and attribute parameters are used to calculate the association probability. Instead of using augmented state vector, we estimate state parameters and attribute parameters separately. Eq.(5) is the estimation equation for attribute parameters, the residual ν a(k) is computed by eq.(6), which is a weighted sum of the residuals of each attributed parameters. The weights β i(k) are computed by eq.(15), which take into account the statistical distance of both state parameters and attribute parameters(eq.16). Simulation is carried out for tracking two targets whose track intersects at a small angle (10.236°) as shown in Fig.1. Table 1 gives the rate of successful tracking by PDA and APPDA. The dotted curves in Fig.(2) are the statistical error in y axis by PDA, the solid curves are the statistical error in y axis by APPDA. In the dotted curve, there is a peak near the intersection of the track, while there is no such peak in the solid curve. That means our method can achieve good performance even at the intersection of targets′ tracks.
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
1999年第4期539-543,共5页
Journal of Northwestern Polytechnical University
基金
国家自然科学基金
关键词
概率数据关联
数据融合
多目标跟踪
特征参数
attribute parameter, probabilistic data association (PDA), data fusion, multiple targets tracking