相位阵列雷达信号目标识别算法的研究

【作者】 郑晓昆；

【导师】 李会方；

【作者基本信息】 西北工业大学 ， 信号与信息处理， 2001， 硕士

【摘要】 本文讨论了传统α－β，α－β－γ滤波器引入采样间隔可变概念后的三种新型采样间隔自适应变化雷达目标跟踪算法，通过仿真，对各算法的滤波性能及效果进行了比较分析，算法中采样间隔的变化与预测误差的平方根或立方根的大小成反比。结果表明，当跟踪目标运动机动性较大时，α－β－γ滤波器跟踪效果好，在同样的跟踪误差水平下其平均采样间隔明显优于α－β滤波，而α－β滤波则更适于跟踪机动性不大的运动目标。另外，当综合考虑采样点数与跟踪误差时，同样使用α－β－γ滤波，采用采样间隔与预测误差立方根大小成反比可获得更好的滤波效果：而同样在使用α－β滤波时，则采用采样间隔与误差平方根成反比时的滤波性能好。 此外，本文还研究了一种基于连续状态Hopfield神经网络的状态估计算法用于目标跟踪，并对该算法中某些参数的选择变化对跟踪滤波效果的影响进行了详细的分析。仿真结果表明在一定条件下，这种新型神经网络目标跟踪算法的滤波性能与Kalman滤波器相仿，但由于其具有很强的并行数据处理能力，因此从计算时间角度讲，较各传统方法更适用于对实时性要求很高的应用领域，如雷达目标跟踪。更多还原

【Abstract】 The effects of incorporating a time-varying update time into three target-tracking algorithms based on the a and a -filters are presented and compared. The update times are chosen according to the inverse of either the square root or cube root of the position residual. Simulation results show that the a 13 v -filter yields larger update times during the manoeuvre than those obtained with the a 13 -filter, as the residuals obtained with the former filter are smaller during the manoeuvre. It is found that the best compromise between the number of updates and estimation errors is obtained when the update time is chosen according to the inverse of the square root of the residual when the a -filter is being used, and the inverse of the cube root of the residual when the a -filter is being used. Besides that, a state estimator based on continuos-state Hopfield net applying to phased array tracking is discussed in this paper, we also show how to chose some of the estimator design parameters. Simulation results show that the new approach performs similar to Kalman filter. Due to the parallel computational mode of neural net, the new approach is more attractive for real time implementation, such as radar target tracking, from the computational time point of view, than classical trackers.更多还原