基于GF-2号影像的森林优势树种分类

利用遥感数据开展森林资源优势树种的分类对森林资源的监测、森林可持续经营及生物多样性研究具有重要意义。研究针对复杂地形区域的破碎化森林,采用高分二号(GF-2)的多光谱影像作为基础数据进行森林优势树种的精细分类。本文以地形复杂、森林破碎化的湖北省竹山县九华山林场为研究对象,采用面向对象分类方法对树种进行精细分类,比较支持向量法、最近邻法(KNN)和随机森林(RF)三种不同分类算法的分类效果。在尺度阈值为30、合并阈值为95时分割的基础上,利用SVM、KNN和RF分类结果和分类精度差异较大。分类精度最高的是SVM分类方法,总体精度为68.52%,Kappa系数为0.62;其次为随机森林分类法,总体精度为60.29%,Kappa系数为0.54;KNN分类方法精度最低,总体精度为59.41%,Kappa系数为0.53。GF-2号数据能满足树种分类基本需求,在复杂地形和景观破碎化地区用支持向量机进行树种的分类精度更高,但仍存在一定的局限性。

Dominant Tree Species Specific Classified by GF-2 Imagery

The use of remote sensing data to classify forest r dominant tree species is of great significance for monitoring forest resources,sustainable forest management and biodiversity research.In order to study the fragmentation forests in complex terrain areas,the multi-spectral imagery of GF-2 is used as the basic data to carry out the fine classification of forest tree species.This paper focuses on the tree species classification in complex terrain area with GF-2 imagery.The study site was located in Jiuhua forestry station in Zhushan County,Hubei province.The object-based classification method was used in research and then the support vector machine(SVM),K nearest neighbors(KNN)and Random forestry(RF)were compared.The results show that,based on the scale threshold(SL)30 and the merge threshold(ML)value of 95,the SVM,KNN and RF classification results and classification accuracy are quite different.The results found that the SVM method had the best classification accuracy with the overall accuracy 68.52%,kappa coefficient was 0.62.The RF method had the second accuracy with the overall accuracy 60.29%,kappa coefficient was 0.54.The KNN had the lowest accuracy.The data of GF-2 can meet the basic requirements of tree species classification.The classification accuracy of tree species in the complex terrain and landscape fragmentation area is higher,but there are still some limitations.