基于各向异性的红外小目标背景预测及分割  

Background prediction and segmentation of infrared small target based on anisotropy

作　　者：郭红伟[1] 赵伶俐[1] 李娟[1] 刘帅[1] 

机构地区：[1]红河学院工学院,云南蒙自661100

出　　处：《激光与红外》2016年第10期1295-1300,共6页Laser & Infrared

基　　金：国家自然科学基金项目(No.41201418;41301442);云南省教育厅科学基金项目(No.2012Y449);红河学院科学基金项目(No.XJ14Y02)资助

摘　　要：红外小目标易淹没在复杂的起伏背景中,为提高目标的检测能力,往往通过抑制背景来增强目标信号.针对各向同性背景在含有较多边缘轮廓的复杂起伏背景预测方法的不足,提出了各向异性的红外背景预测方法,结合目标与背景在局部梯度间的差异,考虑各向异性微分原理,并改进其边缘停止函数,然后利用其两个最小方向值的均值作为背景预测值,并将背景图灰度变换为0 - 255,最后采用恒虚警阈值法对差分图像进行分割处理,达到提取候选目标的目的,降低真实目标的虚警率.实验表明,各向异性取得良好的背景预测效果,而利用恒虚警阈值对差分图像进行分割有效地减小了虚警,提高目标检测率.丗 To improve the detection of infrared small target in the complex background, the target signals are generallyenhanced by restraining the background. For the shortcomings of the isotropic background prediction method,a kind ofanisotropic infrared background prediction method was proposed. According to the difference of local gradient featuresbetween the target and the background, the edge stopping function of anisotropic partial differential equation was improved.Then the mean of the two least direction values was set as the prediction value of the background, and thebackground gray level changed 0 -255. Finally in order to extract the candidate target and reduce the false alarm rateof the target, the difference image was segmented by the method of constant false alarm threshold. Experimental resultsshow that this method has a good background prediction, and it can effectively reduce the false alarm and improve thedetection rate of the target.

关 键 词：各向异性 背景预测 恒虚警阈值 目标分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]