基于SPOT6数据的建筑物提取规则研究  

Research on building extraction rules based on SPOT6 data

作　　者：付盈 国巧真 潘应阳 汪东川 

机构地区：[1]天津城建大学地质与测绘学院,天津300384

出　　处：《国土资源遥感》2017年第3期65-69,共5页Remote Sensing for Land & Resources

基　　金：天津市自然科学基金项目"天津滨海新区地表水环境信息遥感提取与评价方法研究"(编号:13JCQNJC08600);"基于变化轨迹方法的滨海湿地流失累积效应研究"(编号:15JCYBJC23500);国家自然科学基金重点项目"京津唐地区景观格局演变与生态用地流失特征"(编号:41230633)共同资助

摘　　要：针对SPOT6卫星遥感影像,采用基于规则的方法对建筑物进行提取。首先,分析了每种规则属性提取建筑物的效果,在此基础上制定建筑物提取规则;再分别采用K均值聚类法(K-means)、K临近值法(K nearest neighbor,KNN)、支持向量机法(support vector machine,SVM)和神经网络法进行建筑物提取实验,并与基于规则的方法进行对比;最后,对建筑物提取结果进行精度评价。研究表明,基于该规则的建筑物提取精度高于其他方法,在一定程度上缓解了椒盐现象和同谱异物问题,可为今后SPOT6卫星影像更广阔的应用提供一定的技术支持。For SPOT 6 satellite remote sensing image,a method based on rules was used to extract buildings.Firstly,the authors analyzed the extraction effects of every rule attribute and made the rule extract buildings based on the effect. Then the authors compared the methods of K-means clustering,K nearest neighbor( KNN),support vector machine( SVM) and neural network with the method used in this paper during the research. The precision evaluation of building extraction result shows that the accuracy of this method based on rules is higher than that of other methods. This method relieves the problems of the salt and pepper phenomenon and the same spectrum with foreign bodies,and provides some technical support for the wider application of SPOT 6 satellite images in the future.

关 键 词：SPOT6 规则 建筑物提取 KNN SVM 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]