基于自适应背景预测的无源毫米波弱小目标检测  

Small Target Detection in Passive Millimeter-wave Image Based on Adaptive Background Prediction

作　　者：倪炜[1] 张晓丽[1] 胡飞[1] 

机构地区：[1]华中科技大学电子与信息工程系,武汉430074

出　　处：《微波学报》2011年第3期1-4,共4页Journal of Microwaves

基　　金：国家自然科学基金(40906089);中国航天科技集团公司航天科技创新基金重点项目(CASC200902)

摘　　要：在分析毫米波辐射图像特性的基础上,对基于单帧的二维最小均方误差(Two-dimensional LMS)背景预测算法进行改进,充分利用连续帧图像中小目标的运动信息和背景的缓变特性,设计出基于连续帧的TDLMS滤波器。针对TDLMS算法对步长μ选取困难的问题,本文采用变步长TDLMS自适应算法。实验证明,该滤波算法能在两帧图像时达到最优的背景预测效果,有效滤除杂波,提高图像信杂比。According to the characteristics of passive millimeter-wave images,the Two-dimensional Least Mean Square（TDLMS）algorithm is improved,and a TDLMS filter based on consecutive frames is designed,which makes full use of motion information of small target and slow changes of the background in consecutive frames.For the difficult of selecting step-size μ in TDLMS algorithm,adaptive variable step-size algorithm is adopted.Experiments show that this algorithm can achieve the best prediction results of background in two images,removes clutter and improves the image SCR（signal to clutter rate） effectively.

关 键 词：无源毫米波成像 背景预测 二维最小均方差 相邻帧 变步长 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]