摘要
基于土地利用现状图与经验观测,提取标准水稻NDVI时序曲线,利用傅立叶形状描述子计算MODIS-NDVI时序曲线与标准的水稻NDVI时序曲线形状相似性距离,通过样例数据探测未知像元与样本的相似性距离阈值,从而判别双季水稻种植区域。以江汉平原2010年的数据进行实验,证明此方法识别的双季水稻种植区域面积误差为8.6%,总体精度为80%,较为理想。该方法将遥感光谱信息与几何形状的识别相结合,有效减少了个别时段光谱信息误差引起的识别错误,提高了识别水稻种植区域的有效性。
The standard paddy rice NDVI time series curve was extracted according to the present land use map and empirical observation. The shape similarity distance of MODIS-NDVI time series curve compared with standard paddy rice growth curve was calculated by Fourier shape descriptor. And then, the threshold value, which was detected by similarity distance between samples and unknown pixel data, was used to distinguish the double cropping paddy rice planting zones. Validation of the method was conducted in Jianghan Plain in 2010.The results show that the planting zones error of double cropping paddy rice is approximately 8.6%, and the overall accuracy is 80%.This technique could integrate remote sensing spectral information with geometrical shape identification, provide an effective means to reduce identification error caused by spectral information error during certain time intervals, and improve the identification effectiveness of paddy rice planting zones.
出处
《地理空间信息》
2016年第8期56-60,共5页
Geospatial Information
基金
国家自然科学基金资助项目(41371183
41531180)
中央高校基本科研资助项目(CCNU15A02004)
