摘要
根据二维空间内目标作匀速直线运动和匀速圆周运动的特点,在建立目标运动模型和观测模型的基础上采用基于交互多模算法(IMM)的卡尔曼滤波器对机动目标进行跟踪。但由于IMM算法存在大量的递归与矩阵运算,所以实时性降低,这也是工程上难以实现的原因。通过CUDA对算法进行加速实现,在不影响算法性能的基础上实时性大大提高。仿真结果表明,利用该架构实现的IMM算法不仅能够对匀速直线运动和匀速圆周运动的目标进行实时跟踪,而且在运动模型发生变化时,实时性也得到很大提升,滤波误差也比较小。
According to the characteristics of uniform motion in two-dimensional space and on the basis of the target motion model and observation model, the interacting multiple model（IMM） algorithm based Kalman filter is adopted to the maneuvering target tracking. However, there are a lot of recursive and matrix operations in IMM algorithm, so the real time is reduced. This also is the reason that it is hard to achieve in engineering. In this paper, the CUDA is used to speed up the algorithm. The real-time performance is improved greatly on the premise that the arithmetic performance is not affected. Simulation results show that, even if when motion model is changed, the real-time performance can be improved significantly, and the filtering error is also small.
出处
《雷达科学与技术》
2012年第6期656-659,共4页
Radar Science and Technology
关键词
目标跟踪
统一计算设备架构
交互多模
机动
图形处理器
target tracking
compute unified device architecture（CUDA）
interacting multiple model （IMM） algorithm
maneuverability
graphical processing unit（GPU）
