基于BPNN的自适应机动目标跟踪

Adaptive maneuvering target tracking based on back propagation neural network

为了解决现有的机动目标跟踪算法中时延长的问题,提出一种基于反向传播神经网络(BPNN)的自适应目标跟踪算法。从不同运动状态下的观测值中提取特征量,训练BPNN。根据获得的观测值计算得到特征量,将特征量输入到训练好的BPNN中,根据网络输出的运动模型进行滤波更新。仿真结果表明,提出的方法跟踪精度高于经典交互式多模型(IMM)算法,算法运行时间为0.063 5 s,少于IMM算法运行时间0.098 75 s,一定程度上减少了模型决策延迟,使得机动目标跟踪更具实时性。

In order to solve the problem of delay in the existing maneuvering target tracking algorithm,an adaptive target tracking algorithm based on back propagation neural network(BPNN)is proposed in the paper.The characteristic variable is extracted from the observations under different motion states and the BPNN is trained.The characteristic variable is calculated according to the obtained observations and is input of the trained BPNN.The filter update is performed according to the motion model outputted by the network.The simulation results show that the tracking accuracy of the proposed method is higher than that of the classical interacting multiple model(IMM)algorithm.The running time of the proposed algorithm is 0.063 5 s,which is less than the running time of IMM algorithm 0.098 75 s.The proposed algorithm reduces the delay of the model decision and makes the maneuvering target tracking more real-time.