摘要
用高斯混合模型作为角闪烁噪声的近似统计模型,结合Kalman滤波器,提出了一种利用期望最大化(EM)算法抑制角闪烁噪声的预处理方法。首先采用EM算法处理一帧内的原始测量数据,预先得到目标真实位置的最大似然估计,这个预估计量服从渐进高斯分布,且其方差可求。然后将这个预估计量作为Kalman滤波器的输入量进行跟踪滤波,同时将目标位置的预测值作为下一帧EM迭代过程的初始值,进而形成闭环的跟踪滤波结构。仿真结果表明,该方法有效地抑制了角闪烁,使得Kalman滤波算法更加有效,从而提高了目标跟踪的精度。
Choosing Gaussian mixture model as the approximate statistical model of glint noise, and combining with Kalman filter, a new preprocessing method to restrain glint noise using Expectation Maximization （EM） algorithm is proposed. Firstly, using EM algorithm, we can get the pre-MLE （ Maximum Likelihood Estimator） of the true target position from original measurement data in a frame. The estimator obeys asymptotic Gaussian distribution, whose variance is able to be calculated. Then we send the pre-MLE to the Kalman filter as the input, meanwhile make the prediction of target position as the initial value for the EM iterative process in the next frame. It forms a tracking filter with close-loop. Simulation results show that this method is able to restrain glint efficiently, and ensure the efficiency of Kalman filter, which will improve target tracking.
出处
《电光与控制》
北大核心
2009年第5期81-85,共5页
Electronics Optics & Control
基金
国家重大安全基础项目
