采用多伯努利滤波器的过采样点目标检测前跟踪  

Oversampling point target track-before-detect by Multi-Bernoulli filter

作　　者：李淼[1] 龙云利[1] 李骏[1] 安玮[1] 周一宇[1] 

机构地区：[1]国防科技大学电子科学与工程学院,湖南长沙410073

出　　处：《光学精密工程》2015年第12期3446-3455,共10页Optics and Precision Engineering

基　　金：中国博士后科学基金资助项目(No.2013M532167);上海航天科技创新基金资助项目(No.SAST201303)

摘　　要：为了提高过采样体制下点目标的探测能力,提出了基于多伯努利滤波器的过采样点目标检测前跟踪方法。分析了时空过采样体制的成像过程,给出了时空过采样体制下的点目标成像模型,并将过采样观测过程等效建模为多个单采样线阵同时观测的过程。利用提出的模型,将单采样条件下的图像处理方法自然地扩展到过采样条件下。然后,根据等效观测模型,给出了针对过采样点目标的多伯努利预测和更新过程,并用序贯蒙特卡罗的方式实现了该检测前跟踪方法。最后,通过仿真实验验证了本文算法的有效性。实验结果表明:本文提出的方法能够在图像信噪比≥3的条件下,有效检测和跟踪过采样体制下的多个点目标,目标估计误差≤0.25pixel;与单采样检测前跟踪相比,过采样检测前跟踪能够探测目标能量更弱的小目标。因此,本文方法能够满足高精度天基点目标检测跟踪的需求。A oversampling track-before-detect method for point targets was proposed based on the multi-Bernoulli filter to improve the detection performance of an oversampling system.Firstly,the imaging process was analyzed for the temporal-spatial oversampling system,and an imaging model for the point target was established.Then the temporal-spatial oversampling sensor was equivalent to multiple single-sampling sensors.For this,the image processing method used in single-sampling system could be extended to the oversampling system naturally.Furthermore,the prediction step and update step of Multi-Bernoulli filter were given by this proposed model,and the oversampling track-before-detect method was realized by sequential Monte Carlo method.Finally,the simulation experiment was performed to verify the effectiveness of method proposed.The experimental results show that proposed method can effectively detect and track multi point targets at a signal to noise ratio high than0.25 pixel in oversampling system and the error of target state estimation is less than 0.25 pixel.Incomparison with the single-sampling system,the oversampling method can detect and track dimmer targets and meets the requirements of detection and tracking for space point targets.

关 键 词：时空过采样系统 点目标 检测前跟踪 多伯努利滤波器 序贯蒙特卡罗法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]