基于MODIS NDVI时间序列谱匹配的耕地信息提取研究——以浙中地区为例  

Research on cultivated land extraction based on MODIS NDVI time series matching analysis—a case in middle part of Zhejiang

作　　者：王红说[1] 黄敬峰[1] 徐俊锋[1] 彭代亮[1] 

机构地区：[1]浙江大学环境与资源学院农业遥感与信息技术研究所,浙江杭州310029

出　　处：《浙江大学学报（农业与生命科学版）》2008年第3期334-340,共7页Journal of Zhejiang University:Agriculture and Life Sciences

基　　金：国家科技支撑计划资助项目(2006BAD10A01);国家自然科学基金资助项目(40571115/D0119)

摘　　要：基于MODIS NDVI数据和时间序列谱匹配提取季相相似的耕地面积.选取耕地样点像元的时间序列作为标准时间序列谱,其他像元的时间序列作为测试时间序列谱,对匹配系数进行t测验以确定耕地像元.当匹配位置m=0时,在t测验10%显著水平时耕地的匹配精度最高,相关系数r为0.60;而对均方根误差(RMSE)评价图执行非监督分类,获得的耕地匹配精度显著提高,相关系数为0.79,优于t测验方法.该研究证实了基于时间序列谱匹配的耕地信息提取的可行性,同时匹配精度可反映研究区作物季相的一致性信息.MODIS NDVI data and time series matching method were used to extract the area of cultivated land with similar phenology. Time series of sample pixels from cultivated land were selected as standard time series spectrum, and the others were treated as testing time series spectrum to fix cultivated land according to t-test of matching coefficient. When matching position m=0, the matching accuracy at 10% t test level, with the correlation coefficient 0.60, was the highest of all levels;while the matching accuracy got from unsupervised classification of RMSE appraisal map, with higher correlation coefficient 0.79, was better than that of t-test method. The results indicate that cultivated land information extraction based on time series matching method was feasible and the matching accuracy could be used to reflect the consistent information about the phenologies of crop in study area.

关 键 词：NDVI 时间序列谱匹配 交叉相关 匹配精度 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]