基于最小二乘支持向量机的红外小目标检测  

The Detection of Small and Weak Infrared Targets Based on Least Squres Support Vector Machine

作　　者：张学军[1] 李红[1] 邓鹤[1] 

机构地区：[1]华中科技大学数学系,湖北武汉430074

出　　处：《应用数学》2007年第S1期163-167,共5页Mathematica Applicata

摘　　要：红外图像中的小目标由于其特殊性,难于由现有的图像处理知识来识别.支持向量机由于很难实现于像素级的运算而难于推广.本文采用最小二乘支持向量机,对已有图像数据进行分类训练,然后由训练好的向量机来检测红外图像中的小目标.实验证明该方法比支持向量机有明显的速度优势,具有很好的鲁棒性,对于复杂背景下的红外小目标提取十分有效.Owing to the characters of small targets in infrared pictures,it is difficult for them to be distinguished by the current methods of image segmentation.The Support Vector Machine(SVM) is hardly to be extended because of its own limitde ability to the pixel operation. This article adopts the Leaste Squres Support Vector Machine(LSSVM) to classify the existing picture statistic,and then the LSSVM which has been trained can distinguish the targets. The experiments prove this method has an obvious advantage in speed and robustness than the SVM.It is very efficient to collect the small andinfrared targets in complicated background.

关 键 词：最小二乘支持向量机 红外图像 膨账 归一化 

分 类 号：TN215[电子电信—物理电子学]