非高斯非平稳背景的红外弱小目标检测  

Infrared Dim Targets Detection in Non-Gaussian and Non-stationary Background

作　　者：郝红霞[1] 李红[1] 

机构地区：[1]华中科技大学数学系,湖北武汉430070

出　　处：《激光与红外》2006年第7期604-607,共4页Laser & Infrared

基　　金：国防预研基金项目资助(No.51401040104JWO527)

摘　　要：文中利用最小二乘估计对每一像素邻域的相关长度取最优值,并依据相关长度对相关矩阵采取了一种新的处理方法,解决了非平稳图像像素邻域矢量间的相关矩阵不可逆的问题,改进了Chapple和Bertilone构造的基于高斯模型的弱小目标检测方法,从而得出了一种新的红外目标检测算法,适用于非高斯非平稳背景的红外弱小目标检测。该目标检测算法首先采用一个转化函数将图像高斯化,然后求得可逆的相关矩阵,接着求像素邻域矢量的概率密度,最后依据概率密度确定阈值,提取弱小目标。实验结果证实了文中算法的可靠性和可行性。Least-squares estimation is used to every pixel to acquire the optimal correlation length, and a new measure is taken to deal with the correlation matrix to assurance the correlation matrices to be reversible. The Gaussian model -based detection algorithm by Chapple and Bertilone is improved to be a new algorithm, which can be used for the infrared dim targets detection in the non-Gaussian and non-stationary image. Thus the paper puts forth a new infrared targets detection algorithm. The algorithm transforms the original image to be Gaussian, and meanwhile establishes a Gaussian imagery model at first, Then it uses the reversible correlation matrices calculated by the new algorithm to compute the probability intensity, at last, it determines a threshold of the probability intensity to extract the dim target. The experiments on a serial of infrared images show that the new algorithm is reliable and feasible.

关 键 词：非高斯 非平稳 相关 目标检测 

分 类 号：TN391[电子电信—物理电子学]