基于资源三号卫星影像的面向对象分类方法的关键技术研究  

Research on the Key Technologies of Object-oriented Classification Method Based on ZY-3 Satellite Images

作　　者：李宏力 吕立楠 王文钧 王祺 罗建松 LI Hongli;LYU Li′nan;WANG Wenjun;WANG Qi;LUO Jiansong(The Second Geodetic Surveying Brigade,Ministry of Natural Resources,Harbin 150025,China;The Fifth Surveying,Mapping and Geoinformation Engineering Institute of Heilongjiang Province,Harbin 150081,China;Heilongjiang Institute of Geomatics Engineering,Harbin 150081,China)

机构地区：[1]自然资源部第二大地测量队,黑龙江哈尔滨150025 [2]黑龙江省第五测绘地理信息工程院,黑龙江哈尔滨150081 [3]黑龙江地理信息工程院,黑龙江哈尔滨150081

出　　处：《测绘与空间地理信息》2020年第9期155-158,共4页Geomatics & Spatial Information Technology

摘　　要：以资源三号影像为例,利用e Cognition软件,进行面向对象信息提取方法中影像分割尺度选择、最佳波段组合及影像权重、地物提取常用特征等关键技术研究。经过大量试验表明,不同地物的提取需要在不同尺度上借助光谱特征、纹理特征、自定义特征等多种特征才能完成某种地物的提取。资源三号第4波段信息量最大,第1、2波段相关性显著、光谱信息相似,第1、3、4波段组合的OIF指数值最大、含有的信息量最大。The paper takes ZY-3 satellite image as an example,uses the e Cognition software,conducts key technology research of object-oriented information extract method,including the selection of image segmentation scale,optimal band combination,image weighting,and common features of extracting surface features. Experimental results shows that surface feature extraction needs to use spectral feature,textural feature,customized features at different scales. The fourth band of ZY-3 satellite remote sensing image has the maximum information,the correlation between first and second bands is significant,while their spectral information is similar. The combination of first,third,and fourth bands has the maximum information and the the largest OIF index.

关 键 词：面向对象 多尺度分割 OIF指数 特征选择 

分 类 号：P228[天文地球—大地测量学与测量工程]