基于Hyperion数据的南疆环塔里木盆地经济林树种识别敏感波段选择  

Research on the Sensitive Band Selection of Discrimination of Non-timber Product Tree Species Surrounding Tarim Basin in Southern Xinjiang Based on Hyperion Image

作　　者：冯振峰 王振锡[1,2] 岳俊[1,2] 王玲段 

机构地区：[1]新疆农业大学林学与园艺学院,乌鲁木齐830052 [2]新疆教育厅干旱区林业生态与产业技术重点实验室,乌鲁木齐830052

出　　处：《森林工程》2016年第2期7-13,共7页Forest Engineering

基　　金：新疆维吾尔自治区高校科研计划科学研究重点项目(XJEDU2013I16);国防科工局高分专项(95-Y40B02-9001-13/15-01-01);中国博士后科学基金项目(2015M572668XB)

摘　　要：基于Hyperion高光谱遥感影像,以阿克苏红旗坡农场周围的3种经济林树种为研究对象,采用最佳指数法、自适应波段指数法和基于TM影像对应波段划分的波段指数法筛选出4种敏感波段组合,并运用最大似然分类方法对其进行树种识别敏感性检验。研究结果表明:筛选出的4种波段组合对3种经济林树种都有较好的可分性,总体分类精度均达到了75%以上,尤其是基于TM影像对应波段划分的波段指数法筛选的55-33-25和55-158-33组合,分类精度高达86.85%和87.47%,Kappa系数均在0.75以上,可作为南疆环塔里木盆地3种主栽经济林树种大面积遥感识别的依据。In this paper,three kinds of non- timber product tree species surrounding Hongqipo farm in Aksu were studied and the Optimum Index Factor method,Adaptive Band Index method,Band Index method based on the corresponding bands of TM image were used to select 4 kinds of sensitive band combinations based on Hyperion hyperspectral remote sensing image. In addition,the sensitivity of non- timber product tree species discrimination of the selected band combinations was tested by using the Maximum Likelihood method. The results showed that the band combinations selected had good divisibility for the three kinds of main non- timber product tree species and their overall classification accuracy reached more than 75%. Especially for the combinations of 55- 33- 25 and 55- 158- 33 selected by Band Index method based on the corresponding bands of TM image,the Kappa coefficient were more than 0. 75 and the classification accuracy were 86. 85% and 87. 47%,respectively,which can be used as the evidence for large area remote sensing to identify the three kinds of non- timber product tree species surrounding Tarim Basin in southern Xinjiang.

关 键 词：经济林树种 高光谱遥感 数据降维 波段组合 

分 类 号：S771.8[农业科学—森林工程]