在连年种植桉树的区域选取了不同龄期及其生长特性的样本点形成桉树样本集,并以Landsat 8为数据源进行样本集NDVI等8种指数的统计规律性分析,构建了一种基于指数分布规律性的自动阈值决策树分类方法,通过GEE平台将该方法应用于研究区的桉树林分类中,试验表明:1)2013—2019连年种植桉树样本集的多种指数变化具有明显的规律,各种指数每隔3年出现一次极小值,符合桉树的种植-砍伐周期性;2)与随机森林算法分类结果相比,自动阈值决策树的分类结果精度提高了约4%,平均分类总体精度达到0.88,平均kappa系数达到0.83;3)利用谷歌历史影像对自动阈值决策树分类结果进行验证,桉树分布区域重合率达到88.4%。以上结论均表明本文提出的自动阈值决策树分类法能有效实现桉树信息提取。

Eucalyptus of different ages and growth characteristics were selected to form the sample set,and the statistical analysis of the sample set of 8 kinds of NDVI was carried out by using Landsat 8 as the data source.Therefore,an automatic threshold decision tree classification method based on the law of exponential distribution is proposed,which is applied to the Eucalyptus forest classification in the study area through GEE.Results show that:1) from 2013 to 2019,the indexes of the sample set of planting eucalyptus follow certain rules,each index presents a minimum value every three years,which accords with the periodicity of planting and felling of Eucalyptus;2) compared with the classification result of random forest algorithm,the accuracy of classification result of automatic threshold decision tree is improved by about 4%,the average total accuracy of classification is 0.88,the average Kappa Coefficient is 0.83;3) Google historical image is applied to verify the classification result of automatic decision tree,and the coincidence rate of eucalyptus distribution area is 88.4% .All the above results show that the automatic threshold decision tree classification method proposed in this paper can effectively achieve information extraction of Eucalyptus.