摘要
文章提出了一种结合改进的最佳指数法(OIF)和支持向量机(SVM)进行高光谱遥感影像分类新方法。利用本文提出的稳定系数进行波段初选择,根据相关系数选择波段组合生成新影像,并对新影像进行OIF计算,得到OIF值最大的波段组合为最佳波段组合;构建SVM分类器,对最佳波段组合分类;最后将分类结果与其他监督分类方法比较,并在相同核函数下与PCA和SVM结合的方法进行精度比较分析。实验结果表明,本文方法能够有效提取最佳波段组合,在SVM算法下获得较高分类精度。
A novel method was developed to classify hyperspectral remote sensing images based on im- proved Optimum Index Factor （OIF） and Support Vector Machine （SVM） algorithm in the paper. First, a steady coefficient was proposed to select initial bands. Second, some bands according to the correlation be- tween the remaining bands and the initial bands were selected to produce the new image. Third, the OIF was calculated for the new image, and found that the best band combination was with the maximum OIF value. Then, the best band combination was classified by SVM with ＂One-Against-One＂ classification strategy. Finally, the comparison between the accuracies of the classification and other supervised classifi- cations was carried out. The experimental results indicated that the optimal classification bands combina- tion could be effectively obtained with higher accuracy of classification by SVM algorithm.
出处
《测绘科学》
CSCD
北大核心
2014年第11期114-117,66,共5页
Science of Surveying and Mapping
基金
国家973计划重点项目(2010CB731801)
国家自然科学基金项目(61172174)
重大科技专项(2012YQ16018505
2013BAH42F03)
国家海洋局数字海洋科学技术重点实验室开放基金资助(KLDO201307)
关键词
高光谱
最佳指数法
支持向量机
遥感影像分类
hyperspectral
Optimum Index Factor （OIF）
Support Vector Machine （SVM）
remotesensing image classification
