基于SOFM的遥感图像的机场区域变化检测  

SOFM-based Change Detection of Airdrome Area in Remote Sensing Image

作　　者：管新[1] 吴秀清[1] 王建将[1] 

机构地区：[1]中国科学技术大学电子工程与信息科学系,合肥230027

出　　处：《计算机工程与应用》2004年第24期183-185,191,共4页Computer Engineering and Applications

基　　金：国家863高技术研究发展计划项目(编号:2002AA783055)

摘　　要：文章提出了一种新的针对遥感图像机场感兴趣区域变化检测的方法,该方法采用自组织特征映射神经网络对图像进行分类,结合感兴趣区域位置定位结果,利用区域生长分割出感兴趣区域。通过图像间差值,从而获得差异,并利用形状分析的方法对检测出的变化给出了解释。该方法消除了非感兴趣区域变化造成的虚警,克服了多幅图像由于摄取时间和摄取环境的差异造成的影响,提高了检测率。实验表明该文提出的方法具有良好的效果。A new change detection method for Region of Interest(ROI )of airdrome remote sensing image is proposed.This method,which takes advantage of Self-Organizing Feature Map(SOFM)neural network for classification and the ROI's location information,then uses region growing to segment these images,can do well to apart the ROI out of the remote sensing images and get the differences easily from these images.Afterwards it gives a reasonable interpretation for changes by means of shape analysis.This approach not only gets rid of the changes of no interest areas as false alerts,but also avoids the effect of images' difference caused by getting from the different time and the different condi-tions.So it gives good results in experiments.

关 键 词：变化检测 图像分割 神经网络 多边形近似 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]