基于无人机高光谱影像和机器学习的红树林树种精细分类  

Classification of Mangrove Species with UAV Hyperspectral Imagery and Machine Learning Methods

作　　者：姜玉峰 齐建国[1] 陈博伟 闫敏 黄龙吉 张丽 Jiang Yufeng;Qi Jianguo;Chen Bowei;Yan Min;Huang Longji;Zhang Li(Department of Surveying and Mapping,School of Information Science and Engineering,Shandong Agricultural University,Tai'an 271018,China;Key Laboratory of Digital Earth Science,Aerospace Information Research Institute,Chinese Academy of Sciences,Beijing 100094,China;Hai Nan Dong Zhai Gang National Nature Reserve Authority,Haikou 571129,China)

机构地区：[1]山东农业大学信息科学与工程学院测绘系,山东泰安271018 [2]中国科学院空天信息创新研究院,数字地球重点实验室,北京100090 [3]海南东寨港国家级自然保护区管理局,海南海口571129

出　　处：《遥感技术与应用》2021年第6期1416-1424,共9页Remote Sensing Technology and Application

基　　金：中国科学院战略性先导科技专项(A类)(XDA13020506);国家自然科学基金项目(41771392)资助。

摘　　要：利用海南省文昌市清澜港红树林保护区的无人机高光谱影像,采用递归特征消除的随机森林算法(Recursive Feature Elimination-Random Forest,RFE-RF)优选植被光谱特征和纹理特征,通过机器学习中的随机森林(Random Forest,RF)和支持向量机(Support Vector Machine,SVM)算法对研究区内的红树林树种进行精细分类,并对比分析和评价分类模型参数设置对总体精度的影响。结果表明:RF分类方法的总体精度为92.70%、Kappa系数为0.91,与传统的SVM分类方法相比,RF算法均提高了5类树种的生产者精度和使用者精度,能够有效地对红树林树种进行精细分类,可为种植资源规划和生态环境保护等方面提供技术支持。In this paper,we used the UAV hyperspectral images of the mangrove reserve at Qinglan Harbor,Wenchang,Hainan Province,and then preferentially selected vegetation spectral features and texture feature variables using Recursive Feature Elimination-Random Forest(RFE-RF).We further used the Random Forest(RF)and Support Vector Machine(SVM)algorithms to classify the mangrove tree species in the study area,and further the results of the classification model parameters on the overall accuracy were analyzed and evaluated.The results showed that the overall accuracy of RF classification was 92.70% and the Kappa coefficient was 0.91.Compared with the traditional SVM classification method,RF improved the producer accuracy and user accuracy of five types of tree species,which could effectively classify mangrove tree species and provide technical support for germplasm resource planning and ecological environmental protection.

关 键 词：机器学习 随机森林 高光谱 特征提取 精细分类 

分 类 号：P23[天文地球—摄影测量与遥感] TP79[天文地球—测绘科学与技术]