基于SPOT6数据的建筑物提取规则研究

发布时间：2018-03-24 20:37

  本文选题：SPOT　切入点：规则　出处：《国土资源遥感》2017年03期

【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。
[Abstract]:In view of SPOT6 satellite remote sensing image, a rule-based method is used to extract buildings. Firstly, the effect of each rule attribute extraction on buildings is analyzed, and then the building extraction rules are established. Then K-means clustering method and K-means-K approach method, support vector machine method, support vector machine method and neural network method are used to carry out the experiment of building extraction, and compared with the rule-based method. Finally, the method of support vector machine (SVM) and neural network (NN) are compared with the rule-based method. The research shows that the precision of building extraction based on this rule is higher than that of other methods, which alleviates the salt and pepper phenomenon and the problem of isospectral foreign body to some extent. It can provide certain technical support for wider application of SPOT6 satellite image in the future.
【作者单位】： 天津城建大学地质与测绘学院;
【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
【分类号】：P237;TP751
，

本文编号：1659931