基于Contourlet变换和Facet模型的红外小目标检测方法  

Method of infrared small target detection based on Contourlet transform and Facet model

作　　者：卢瑞涛[1] 黄新生[1] 徐婉莹[1] 

机构地区：[1]国防科学技术大学机电工程与自动化学院,湖南长沙410073

出　　处：《红外与激光工程》2013年第8期2281-2287,共7页Infrared and Laser Engineering

基　　金：实验室青年教师创新研究项目(LWAS-QNJS-2011-05)

摘　　要：针对存在复杂背景干扰和噪声情况下的红外弱小目标检测问题,提出了一种基于循环平移Contourlet变换和Facet模型多向梯度特性的检测方法。首先通过循环平移Contourlet变换,利用硬阈值对图像进行去噪,提高图像的信噪比和平滑性;然后设计了一种基于Facet模型多向梯度特性的中值滤波器,对去噪后的图像进行滤波,有效地抑制复杂背景和噪声;其次采用两级最大类间方差算法对滤波后的图像进行分割;最后根据相邻帧候选目标的位置和速度关系进一步检测弱小目标。实验证明,这种算法抗噪性强,对包含强纹理结构的复杂背景具有良好的抑制作用,能够有效地检测出弱小目标。It is a problem to detect small moving target in infrared images, especially under the circumstances of complex background and noise. Now aiming at this problem, a novel method based on cycle spinning Contourlet transform and the multi-orientation gradient character of Facet model was proposed. Firstly, through cycle spinning Contourlet transform, the infrared image was decomposed so that the noise was removed and the signal-to-noise ratio and smoothness were increased. Then, a median filter based on the multi-orientation gradient character of facet model was designed to filter the reconstructed image, so the complex background and noise were restrained effectively. After that, the algorithm of 2- level maximum between-cluster variance was used to segment the filtered image. Finally, further detection of the small target was made according to the relationship of position and velocity in images. The results prove that the method concerned in this paper can restrain the complex background with strong clutter to good effect and can test small target efficiently.

关 键 词：红外小目标 循环平移Contourlet变换 FACET模型 中值滤波 阈值分割 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]