摘要
针对遥感技术在土地覆盖分类中应用的重要性,以Landsat TM影像为数据源,选择传统的最大似然分类器、最小距离分类器和新兴的支持向量机分类器、以及面向对象分类方法,设计不同的分类判据特征组合,对不同分类器、不同分类特征组合的精度进行比较分析。结果表明,最大似然分类器、支持向量机分类器都具有良好的分类效果,综合使用多种特征作为输入可以提高分类精度,适合于研究区域土地覆盖分类。
Remote sensing images play important roles for urban land cover classification and change analysis. Using Landsat TM image as data source, traditional maximum likelihood classifier, minimum distance classifier and novel support vector machine are used to classify urban land cover from pixel level, and object-oriented classification method is also experimented from object level. Based on an overall analysis to the performance of various classifiers and the impacts of input features on classification accuracy, it is shown that MLC and SVM are effective to urban land cover classificaion, and the combination of multiple＂ features can improve classification accuracy. The performance of object-oriented method is close to MLC and SVM.
出处
《辽宁工程技术大学学报（自然科学版）》
CAS
北大核心
2010年第4期567-570,共4页
Journal of Liaoning Technical University (Natural Science)
基金
国土环境与灾害监测国家测绘局重点试验室开放式基金资助项目(LEDM2009C04)
龙岩市科技局计划项目(2009LY71)
关键词
土地覆盖
分类
支持向量机:面向对象影像分类
land cover
classification
support vector machine
object-oriented classification
