摘要
基于Hyperion高光谱影像,对提取植被覆盖度的传统像元二分法进行了改进,提出通过地物分类来提高植被覆盖度提取精度的算法.该算法先对研究区进行分类,在较高分类精度的基础上,结合不同地物的NDVI频率累积图和实际情况得到各类地物的植被覆盖度,最后得到研究区域的整体植被覆盖度.结果表明,经支持向量机分类的总体精度为83.2%,Kappa系数为0.710;相同NDVI值,林地的植被覆盖度>农田的植被覆盖度>草地的植被覆盖度,实验结果与实际基本相符;改进的像元二分算法改善了传统像元二分法中存在的水体、裸地非植被覆盖区得到非0的植被覆盖度和高植被覆盖地区检测灵敏度下降等问题,丰富了传统像元二分法的细节信息,得到更符合实际的植被覆盖度.
This paper improved the traditional method for extraction of vegetation cover,pixel dichotomy,based on the Hyperion hyperspectral imaging,and proposed the algorithm that enhanced the accuracy of such extractionby categorizing objects.The method first categorized the research area.On the basis of higher categorizing accuracy,it obtained the vegetation cover of different surface features combined with the NDVI cumulative frequency diagram and the actual situation.Finally,the overall vegetation coverage of the study area was obtained.The result shows that the overall accuracy according to the classification of the support vector machine(SVM)is 83.2%and the Kappa coefficient is 0.710;based on the same normalized difference vegetation index(NDVI),the forest vegetation coverage>the farmland vegetation coverage>the grassland vegetation coverage,which is basically insistent with the actual situation.And the existing problems in traditional pixel dichotomy that the water and non-vegetation covered areas in the bare land could also produce non-zero vegetation coverage results and areas of higher vegetation cover were found to have lower detection sensitivity are improved by improved pixel dichotomyalgorithm,the improved method enriches the details of the traditional pixel dichotomy,researchers could achieve a more realistic vegetation coverage.
作者
张贵成
林志垒
ZHANG Guicheng;LIN Zhilei(Fujian Provincial Engineering Research Center for Monitoring and Assessing Terrestrial Disasters,Fujian Normal University,Fuzhou 350007,China;School of Geographical Sciences,Fujian Normal University,Fuzhou 350007,China)
出处
《福建师范大学学报（自然科学版）》
CAS
北大核心
2020年第1期79-86,95,共9页
Journal of Fujian Normal University：Natural Science Edition
基金
国家重点研发计划项目(2016YFC0502901)
福建省自然科学基金资助项目(2017J01463)
关键词
植被覆盖度
像元二分法
高光谱
分类
vegetation coverage
pixel dichotomy
hyperspectral
classification
