基于SVM雷达卧龙大熊猫栖息地森林成图
发布时间：2018-09-01 16:24
【摘要】：卧龙自然保护区(世界自然遗产地)是大熊猫最主要的栖息地之一。结合雷达遥感全天时、全天候观测优势,以及森林覆盖对栖息地生境评价的重要性,开展多时相、双极化雷达数据森林精细成图研究就显得尤为重要。本研究首先对雷达数据进行辐射地形校正;然后选用5个时相ALOS PALSAR数据,采用支持向量机(support vector machine,SVM)方法进行森林精细成图。研究选取了5个多时相、双极化典型特征信息参与初始训练和分类,即HH_m,HV_m,TSD,HH_m-HV_m和HH_m/HV_m;接着通过对不同信息组合分类精度的试验与对比,获取了最优特征组合HH_m,HV_m,TSD,HH_m-HV_m。对应分类总体精度、森林及非森林类别用户精度分别为86.90%,82.34%和92.83%,显著优于单时相单极化数据分类结果(分类总体精度55.47%)。研究结果验证了多时相、双极化雷达遥感数据在自然遗产地森林精细成图中的有效性,并揭示了雷达遥感在多云多雨地区生境监测与评价中的潜力与应用价值。
[Abstract]:Wolong Nature Reserve is one of the most important habitats for giant pandas. Combined with the advantages of radar remote sensing all day, all weather observation and the importance of forest cover to habitat evaluation, it is very important to study the precise mapping of forest with multi-phase and dual-polarization radar data. In this study, the radar data were first corrected by radiometric topography, and then five time-phase ALOS PALSAR data were selected, and the support vector machine (support vector machine,SVM) method was used for fine forest mapping. In this study, five multi-phase and two-polarization typical feature information were selected to participate in the initial training and classification. Then, HH_m,HV_m,TSD,HH_m-HV_m and HH_m/HV_m; obtained the optimal feature combination HH_m,HV_m,TSD,HH_m-HV_m. by the experiment and comparison of the classification accuracy of different information combinations. The accuracy of forest and non-forest users was 86.90% and 92.83% respectively, which was significantly better than that of single time and single polarization data (total classification accuracy was 55.47%). The results verify the validity of multi-temporal and dual-polarization radar remote sensing data in the fine forest mapping of natural heritage areas, and reveal the potential and application value of radar remote sensing in monitoring and evaluation of habitats in cloudy and rainy areas.
【作者单位】： 中国科学院遥感与数字地球研究所数字地球重点实验室;中国科学院大学;联合国教科文组织国际自然与文化遗产空间技术中心;山东科技大学测绘科学与工程学院;
【基金】：国家国际科技合作专项项目“全球变化对世界遗产影响空间精细观测与认知”(编号:2013DFG21640) 中国科学院“百人计划”项目“雷达遥感考古机理与方法示范研究”(编号:Y5YR0300QM)共同资助
【分类号】：TN957.52
[Abstract]:Wolong Nature Reserve is one of the most important habitats for giant pandas. Combined with the advantages of radar remote sensing all day, all weather observation and the importance of forest cover to habitat evaluation, it is very important to study the precise mapping of forest with multi-phase and dual-polarization radar data. In this study, the radar data were first corrected by radiometric topography, and then five time-phase ALOS PALSAR data were selected, and the support vector machine (support vector machine,SVM) method was used for fine forest mapping. In this study, five multi-phase and two-polarization typical feature information were selected to participate in the initial training and classification. Then, HH_m,HV_m,TSD,HH_m-HV_m and HH_m/HV_m; obtained the optimal feature combination HH_m,HV_m,TSD,HH_m-HV_m. by the experiment and comparison of the classification accuracy of different information combinations. The accuracy of forest and non-forest users was 86.90% and 92.83% respectively, which was significantly better than that of single time and single polarization data (total classification accuracy was 55.47%). The results verify the validity of multi-temporal and dual-polarization radar remote sensing data in the fine forest mapping of natural heritage areas, and reveal the potential and application value of radar remote sensing in monitoring and evaluation of habitats in cloudy and rainy areas.
【作者单位】： 中国科学院遥感与数字地球研究所数字地球重点实验室;中国科学院大学;联合国教科文组织国际自然与文化遗产空间技术中心;山东科技大学测绘科学与工程学院;
【基金】：国家国际科技合作专项项目“全球变化对世界遗产影响空间精细观测与认知”(编号:2013DFG21640) 中国科学院“百人计划”项目“雷达遥感考古机理与方法示范研究”(编号:Y5YR0300QM)共同资助
【分类号】：TN957.52
【相似文献】
相关期刊论文 前10条
1 刘国梁;毫米波双极化传输技术的研究[J];电信科学;1988年06期
2 潘雪明,王泽美,焦永昌;一种槽耦合的双频双极化微带天*线[J];雷达科学与技术;2005年04期
3 康雪姣;张金生;;一种双频双极化3G天线的研究[J];微计算机信息;2008年32期
4 金擰民;;利用煇卡，
本文编号：2217731
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2217731.html
