摘要
利用多时相的Landsat TM/ETM+和DEM数据,分别采用直方图匹配(HM)和准不变特征点(PIFs)方法对影像序列进行相对辐射校正,减少了影像的光照和大气条件在时间上的不确定性,提高了归一化植被指数(NDVI)和地表水分指数(LSWI)的计算精度。根据水稻在不同生育期表现出的生理特征,基于LSWI和NDVI时间序列及高程特征,采用二叉树方法提取了浙江省金华市水田信息。经过验证,在空间上水田信息的提取精度达到92.3%,在县域尺度上提取面积与统计年鉴具有0.97的相关度。
Rice is one of the most important food crops in China. Timely and accurately acquiring the area and spatial distribution of paddy fields is significant to the crop yield estimates, food security and water resources management and so on. In this paper, the multi-temporal Landsat TM/ETM-＋ images and DEM are used and histogram matching and pseudo invariant feature spot methods are employed respectively to make relative radiometric correction for the TM/ETM＋ image sequence. It reduces the uncertainty of the lighting and atmospheric conditions in the temporal phase for the images, and improves the accuracy of the NDVI and LSWI. Firstly, the cultivated land of study area is extracted quantificationally according to multi-temporal NDVI, then based on the physiological characteristics of rice in different growth stages, the paddy field containing partial dry land is extrac- ted by defining the threshold of LSWI in June and August. On the basis,according to the feature that the NDVI of double crop- ping rice has two peaks which appear in June and September respectively and the NDVI of single cropping rice has one peak in August, the paddy field is extracted more accurately. The results show that the precisions of paddy field extraction are as high as 92. 3%. Besides, the correlation coefficient between extraction area and statistical yearbook is 0. 97 in county scale.
出处
《地理与地理信息科学》
CSCD
北大核心
2015年第3期32-37,F0002,共7页
Geography and Geo-Information Science
基金
国家自然科学基金项目“遥感影像小支撑规范化去卷积快速算法研究”(41371331)
