基于DTW距离的时序相似性方法提取水稻遥感信息——以泰国为例  

Extraction of Paddy Rice Area Using a DTW Distance Based Similarity Measure

作　　者：管续栋 黄翀[1] 刘高焕[1] 徐增让[1] 刘庆生[1] 

机构地区：[1]中国科学院地理科学与资源研究所资源与环境信息系统国家重点实验室,北京100101 [2]山东师范大学人口资源与环境学院,济南250014

出　　处：《资源科学》2014年第2期267-272,共6页Resources Science

基　　金：中国科学院重点部署项目(编号:KZZD-EW-08);中国科学院地理科学与资源研究所"一三五"战略科技计划项目(编号:2012SJ008);国家科技部基础性工作专项(编号:2008FY110300)

摘　　要：热带季风区多云多雨的天气条件一直是多光谱遥感探测地表信息的难点之一。本文针对东南亚地区多雨多云的复杂天气条件以及水稻种植灵活的特点,利用MODIS时间序列数据,提出一种基于动态时间弯曲(DTW)距离的相似性判别的土地覆盖分类方法,对泰国东北部地区单、双季稻种植面积进行了遥感提取研究。针对研究区雨季遥感影像像元受到云覆盖影响严重,使用替换法去云,结合S-G滤波方法对计算得到的MODIS09A1数据的NDVI时序数据去噪,再采用DTW距离相似性方法逐像元比较与标准NDVI时间序列的时序相似性,将不同类型所得NDVI相似性值作为模糊分类隶属度参考值对泰国东北部地区单季稻、双季稻进行分类提取面积。最后结合野外采样数据、Google Earth高清遥感影像进行精度验证。结果表明,该方法能够用于针对东南亚多雨多云区水稻种植面积大范围监测。MODIS time-series data allows for the possibility of extracting paddy rice area in a timely and accurate way because of its high spectrum and time resolution. Here, we focused on the complicated weather conditions and flexibility of rice cultivation using MODIS time series data and propose a DTW distance-based similarity measure for land coverage classification. We extracted the single cropping rice and double cropping rice area of northeastern Thailand. In view of cloud on the RS image of the research area during monsoon, we first used a replacement method processing cloud remove, and combine with S-G filtering method denoise the NDVI time-series computed by MODIS09A1 data. Then we used the DTW distance similarity measurement to compare each pixels＇ NDVI time series with the standard NDVI time series, and calculate the similarity measure value of each class as the reference value in a fuzzy classification membership. Finally, we obtained the area of single and double rice in northeast Thailand and combined field sampling data, Google Earth high-resolution images to verify the spatial accuracy of the outcome. The result of this research shows that this method can be used for a wide range of rice area monitoring using MODIS time-series data.

关 键 词：MODIS时序数据 DTW 模糊分类 泰国 水稻 

分 类 号：S511[农业科学—作物学] S127[天文地球—摄影测量与遥感] P237[天文地球—测绘科学与技术]