论文部分内容阅读
提出一种改进的双通道交互多模型(IMM)算法,使用x和y维残差的边缘分布独立表征预测误差的似然函数,使滤波模型在不同维度与目标运动中分别匹配.首先分析了残差-似然函数-模型概率更新之间的映射关系,证明算法的可行性和有效性,然后引入图形处理器(GPU)并行运算平台,设计了通道级-模型级嵌入式并行方案,提高算法的实时性.仿真表明:相比于传统基于CPU平台下的IMM算法,该算法预测误差与真实噪声似然分布契合度更高、实时性更强、跟踪误差更小.
An improved two-channel interactive multiple model (IMM) algorithm is proposed, which uses the marginal distribution of x and y dimensional residuals to independently represent the likelihood function of prediction error, so that the filtering model matches with the target motion in different dimensions respectively. Residual-likelihood function-model probability update, and proves the feasibility and effectiveness of the algorithm. Then, we introduce the parallel computing platform of graphics processor (GPU) to design a channel-level model-level embedded parallel scheme and improve The simulation results show that compared with the traditional IMM algorithm based on CPU platform, this algorithm has a better fit between the prediction error and the real noise likelihood distribution, the better real-time performance and the smaller tracking error.