| 动态时间弯曲技术支持下时序NDVI数据的耕地分布信息提取 |
| |
| 引用本文： | 孙越凡,钟礼山,程亮,李满春.动态时间弯曲技术支持下时序NDVI数据的耕地分布信息提取[J].资源科学,2014,36(9):1977-1984. |
| |
| 作者姓名： | 孙越凡 钟礼山 程亮 李满春 |
| |
| 作者单位： | 南京大学地理信息科学系, 南京 210023;南京大学地理信息科学系, 南京 210023;江苏省地理信息技术重点实验室, 南京 210023;南京大学地理信息科学系, 南京 210023;江苏省地理信息技术重点实验室, 南京 210023;南京大学地理信息科学系, 南京 210023;江苏省地理信息技术重点实验室, 南京 210023 |
| |
| 基金项目： | 国家科技支撑项目（编号：2012BAH28B02）；南京大学大学生创新训练项目（编号：XY1310284018）。 |
| |
| 摘 要： | 我国耕地保护形势十分严峻，及时获取耕地现状信息对耕地保护工作的开展具有重要参考意义。文章探讨了动态时间弯曲技术在时序NDVI数据的耕地分布信息提取中的可行性，并以江苏省苏南地区为实验区，用2010年MODIS NDVI时间序列数据进行耕地提取实验。首先使用频域低通滤波对时序NDVI数据进行重构，抑制大气、光照等噪声的影响，接着选取耕地训练样本，构建耕地参考时间序列，然后根据动态时间弯曲（DTW）方法计算待分类像元时间序列与耕地参考时间序列的相似性程度，对计算结果使用阈值分割，得到研究区耕地信息。实验中耕地提取结果正确率达84.08%，完整率达85.68%，证实了该方法的适用性。研究表明，基于时序NDVI数据相似性分析可以快速有效提取大面积耕地信息，为农业政策的制定提供参考依据。
|
| 关 键 词： | 耕地分布 NDVI 时间序列 相似性分析 DTW |
Cropland Information Extraction from Time Series NDVI Data Using Dynamic Time Warp |
| |
| SUN Yuefan,ZHONG Lishan,CHENG Liang and LI Manchun.Cropland Information Extraction from Time Series NDVI Data Using Dynamic Time Warp[J].Resources Science,2014,36(9):1977-1984. |
| |
| Authors: | SUN Yuefan ZHONG Lishan CHENG Liang and LI Manchun |
| |
| Institution: | Department of Geographic Information Science, Nanjing University, Nanjing 210023, China;Department of Geographic Information Science, Nanjing University, Nanjing 210023, China;Jiangsu Provincial Key Laboratory of Geographic Information Science and Technology, Nanjing 210023, China;Department of Geographic Information Science, Nanjing University, Nanjing 210023, China;Jiangsu Provincial Key Laboratory of Geographic Information Science and Technology, Nanjing 210023, China;Department of Geographic Information Science, Nanjing University, Nanjing 210023, China;Jiangsu Provincial Key Laboratory of Geographic Information Science and Technology, Nanjing 210023, China |
| |
| Abstract: | The protection of cropland resources in China is paramount and information about cropland resources is urgently needed. The accurate extraction of cropland information from remotely sensed data is quick and cheap, but remains challenging. Due to artificial cultivation, regular changes in cropland from multi-temporal remotely sensed images provides a possible cue for information extraction but is influenced by many factors such as geographic distribution, climate, crop types and scale. Here, we use time series NDVI datasets for cropland information extraction. In time series analysis, dynamic time warping(DTW)is an algorithm for measuring similarity between two temporal sequences which may vary in time. Similarities in patterns are detected using DTW, even if one curve is different from the other, to a certain extent. A low-pass filtering in frequency domain was first applied to re-establish the NDVI time series and the impact of light and atmospheric noise is then weakened. A couple of training samples were selected as reference time curves of cropland. For each to-be-classified pixel, a quantitative similarity between its time-curve and reference time-curve was calculated using DTW and cropland distribution information was extracted after a segmentation process. An experiment in southern regions of Jiangsu province was conducted and a producer accuracy of 85.68% and user accuracy of 84.08% was achieved. |
| |
| Keywords: | cropland distribution NDVI time series similarity analysis DTW |
|
| 点击此处可从《资源科学》浏览原始摘要信息 |
| 点击此处可从《资源科学》下载免费的PDF全文 |
|
