云背景下红外小目标检测的分形方法  

Fractal method of infrared small targets under cloud background

作　　者：郭同健[1] 高慧斌[1] 宋立维[1] 余毅[1] 

机构地区：[1]中国科学院长春光学精密机械与物理研究所,吉林长春130033

出　　处：《激光与红外》2014年第11期1278-1281,共4页Laser & Infrared

基　　金：国家863计划资助项目(No.2010AA7034320)资助

摘　　要：针对云背景下红外小目标检测对背景起伏依赖性强的问题,提出采用分形方法实现不同起伏背景的小目标检测。首先利用分数布朗随机场模型对实际云背景图像建模得到分形曲线,然后通过分析小目标存在对云背景分形特性的影响,总结出云背景区域含小目标时其分形曲线面积增大的规律,据此提出了基于分形曲线面积差量的小目标检测方法。实验结果表明:在虚警率为10-3时,分形方法对弱起伏和中等起伏云背景下红外小目标检测概率达到1.0,对强起伏云背景下红外小目标检测概率达到0.6,较现有其他一些方法具有更好的检测效果和背景适应能力。To solve the problem that the detection of infrared small targets under cloud background is strongly depend on the fluctuation of cloud,the fractal method is proposed. Firstly, the actual cloud background image is modeled by the fractional Brownian random field, and the fractal curve is obtained. Then the influence of small targets on the frac- tal characteristics of cloud background is analyzed ,and the conclusion is draw that the fractal curve area will increase if small targets exist under cloud background. Finally, the approach to detect small targets is proposed, which is based on the fraetal curve area dispersion. The results of experiments show that the target detection probability is up to 1.0 under weak and median cloud background and 0. 6 under strong cloud background through fractal method when false alarm probability is 10^-3 ,which reflects the fractal method is more detective and adaptive to various cloud background than existing methods.

关 键 词：小目标检测 分形 云背景 图像处理 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]