一种云天背景下单帧红外弱小目标检测新方法  

New method of infrared small target detection of single frame under cloudy sky background

作　　者：张彦峰[1] 何佩琨[1] 张辉[1] 

机构地区：[1]北京理工大学信息与电子学院,北京100081

出　　处：《光学技术》2012年第6期745-750,共6页Optical Technique

基　　金：总装预研项目(51301050106)

摘　　要：针对复杂背景下单帧红外图像中的弱小目标检测问题,提出了一种级联的改进Asymmetric AdaBoost算法。该算法通过提取红外小目标梯度对比度、邻域灰度分布等多个有效特征,充分利用多个红外特征的互补作用,对Asym-metric AdaBoost在仿真中出现的问题进行了相应的改进。通过对实测数据进行仿真,结果表明该方法达到了高检测概率和低虚警率,能有效地检测红外弱小目标。A^tract： Aim at the detection of infrared small target under complicated background, a kind of modified Asymmetric AdaBoost arithmetic used in the cascade is proposed. Through picking up infrared character, grad contrast, distribution of gray level in the neighborhood, etc, it utilizes complementarities between each infrared character fully, and modifies its problem in the stimulation. Simulation result indicates that it can attain high detection ratio and lower false alarm ratio and detect infrared target with high efficiency. Key words： AdaBoost; classifier; ensemble of classifiers; Asymmetric AdaBoost

关 键 词：ADABOOST分类器 分类器集 ASYMMETRIC ADABOOST 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]