基于HMM-PNN模型的助推段目标类型识别

Trajectory classification based on HMM-PNN model in boost phase

针对卫星视野下导弹目标的识别问题,将经典隐马尔科夫模型(HMM)识别算法应用在助推段目标类型识别上并加以改进。首先,分析了通用弹道助推段运动特性,确定了不同射程导弹的分类依据。其次,针对HMM模型时序特性差异较小而引起的识别率低的问题,引入概率神经网络(PNN)与HMM模型相结合的结构算法,该方法整合了HMM模型的时间序列数据处理能力和PNN的自学习能力、贝叶斯决策理论,对不同射程导弹目标实现了分类识别。仿真实验结果表明该算法是一种有效的导弹目标识别算法,识别率优于传统的HMM模型方法,误判率较低,且易于工程实现。

An algorithm of trajectory classification based on classical Hidden Markov Model for early warning satellite system is established and improved in boost phase.First,the classification basis established by analyzing the time sequence property of the normal boost trajectory.Then,an structure algorithm combined HMM model with PNN model is introduced to solve the question of low rate of recognition caused by the less obvious difference among the trajectory,which integrated the capacity of time sequence processing for HMM and the Bayesian decision theory for PNN.Feasibility and effectiveness of the algorithm is verified by simulation experiments.Analysis of the result shows that its performance is remarkably improved compared to traditional HMM method,with its implementation suitable for engineering fulfillment.