摘要
提出一种使用人工神经网络技术来估计红外图像背景的快速算法,并利用红外图像中弱小目标的特性来构建目标模型,采用中心重合的大、小两个窗口,用大窗口的外层来估计目标周围的背景,即隐含层第一个结点的输出值,大窗口内的小窗口则是用来估计中心像素的特性,即隐含层第二个结点的输出值,用隐含层第二个结点减去第一个结点的差的大小来判断中心像素是属于目标还是背景,差值越大输出值越大。采用该思想训练网络权值,可以更好地检测真目标,剔除虚假目标。
A fast algorithm is proposed based on artificial neural network to estimate the background of infrared images,and a target model is constructed using the characteristics of small infrared targets.A big and a small windows sharing the same centre are created.The background is computed with the pixels between the two windows,which is taken as the first node output of hidden layer.The centre characteristics are also calculated with the pixels inside the small one,which are taken as the second node output of hidden layer.The first node is subtracted from the second one,and the result is used to estimate whether the centre pixel is the target or not.The bigger result makes the output bigger.The proposed target model can detect real targets and delete false targets much effectively.
出处
《控制工程》
CSCD
北大核心
2010年第5期611-613,共3页
Control Engineering of China
基金
国家重点基础研究发展计划(973计划)资助项目(2010CB731800)
中国科学院百人计划课题资助
关键词
红外图像
小目标检测
背景估计
人工神经网络
infrared images
small targets detection
background estimating
artificial neural network
作者简介
焦建彬（1967-），男，河北安国人，教授，博士生导师，主要从事模式识别与图像处理等方面的教学与科研工作。
