基于过渡区特征的全色遥感图像冰雪识别  

Recognition of ice or snow for panchromatic remote sensing image based on transition region feature

作　　者：陈婷 汪爱华 王智勇 

机构地区：[1]北京宇视蓝图信息技术有限公司,北京100096 [2]二十一世纪空间技术应用股份有限公司,北京100096

出　　处：《国土资源遥感》2013年第2期27-32,共6页Remote Sensing for Land & Resources

基　　金：北京市科技计划项目(编号:Z111101061710001);国家科技支撑计划项目(编号:2011BAH23B01)共同资助

摘　　要：针对全色图像的冰雪识别问题,以过渡区理论为基础,提出基于过渡区特征的冰雪识别方法。首先利用K-均值聚类方法分离云、雪与其他地物,再通过SUSAN(smallest univalue segment assimilating nucleus)边缘检测提取过渡区图像;然后设立描述过渡区大小的特征量——厚度,并与过渡区的均值和方差特征量组合成特征向量,用以分析过渡区特征,识别具有冰雪过渡区特征的点,构成冰雪边界;最后经过边界生长和区域填充,实现冰雪识别。以"北京一号"小卫星全色图像为遥感数据源,应用该方法及过渡区阈值法、迭代自组织法和面向对象纹理分析法分别提取冰雪覆盖区。该方法的冰雪识别精度达到97.39%,明显高于其他识别方法,表明该方法能获得较高精度的冰雪识别结果和丰富的边缘信息,可为云雪分离及雪线提取等方法研究提供参考。In order to accurately identify ice or snow and obtain the range, this paper presents a new recognition method based on transition region feature for high spatial resolution panchromatic remote sensing imagery. Firstly, the high reflection region including snow or ice and cloud was extracted by K- means cluster analysis. Secondly, the transition region was segmented by SUSAN edge detection. Then, the average, variance and thickness were chosen as the transition region feature vectors to differentiate ice or snow pixels as the target boundary. Finally, the snow or ice area was obtained by edge growing and region filling The ＂Beijing- 1＂ high spatial resolution panchromatic remote sensing image was selected to identify the ice or snow area by transition region feature, and the recognition precision reached 97.39%. A comparison of the experimental results with those of other methods shows that the accuracy of the transition region feature analysis is obviously improved. The application analysis indicates that the method of ice and snow recognition based on transition region feature can obtain higher precision of results and more details of edges, and can also provide the references for separating the cloud and snow and extracting the snow line.

关 键 词：冰雪识别 过渡区 过渡区特征 全色遥感图像 

分 类 号：TP237[自动化与计算机技术—检测技术与自动化装置] TP751[自动化与计算机技术—控制科学与工程]