摘要
快速、无损、准确地监测水稻穗期氮素状况,对于诊断水稻生殖生长特征、提高氮肥运筹水平具有重要意义。本研究在浙江省海宁市晚稻试验点进行田间取样试验,并获取同时期CBERS-1遥感数据,分析了试验点晚稻穗期叶片氮素与CBERS-1影像冠层光谱信息之间的关系。结果表明,水稻穗期叶片氮素含量与同期CBERS-1影像的光谱信息NDVI之间有良好的相关性,可以建立水稻穗期叶片氮素含量反演的相关统计模型。但由于遥感影像特征与水稻穗期叶片氮素含量之间存在较复杂的非线性关系,因此统计模型反演精度不够理想。因而,又尝试运用BP人工神经网络方法来反演水稻穗期叶片氮素含量,发现BP人工神经网络模型具有很强的非线性拟合能力,与统计模型相比,其水稻穗期叶片氮素含量的反演精度有显著提高。由此表明,CBERS-1遥感影像技术与BP人工神经网络方法结合可以对水稻穗期叶片氮素含量进行建模并反演,能够在较大的范围里估测水稻的氮素营养状况。
Non-destructive, rapid and accurate monitoring of rice nitrogen nutrition at spiking stage is significant in estimating rice reproductive growth, enhancing nitrogen management and use efficiency. In this study, the field experiments were carried out at studied fields of Haining City, Zhejiang Province in 2007, according to the pass time of CBERS-1 satellite. Correlation analyses between rice leaf nitrogen and rice canopy spectrum information from the CBERS-1 image were made to all study fields. The leaf nitrogen content at spiking stage had good correlation with the spectrum parameters NDVI. However, due to rather complicated non-linear relations existed between image features and leaf nitrogen content at spiking stage, the results of leaf nitrogen content retrieved from the statistic model was not so ideal. For this reason, an artificial neural network model （BP model） was constructed and applied in the retrieval of leaf nitrogen content at spiking stage. Due to its superior ability for solving the non-linear problem, the BP model provided a much better accuracy in retrieval of leaf nitrogen content compared with the statistic model. It is feasible to use remote sensing technology combined with BP model to predict the nitrogen content of rice leaf at large scale at spiking stage.
出处
《核农学报》
CAS
CSCD
北大核心
2009年第3期364-368,共5页
Journal of Nuclear Agricultural Sciences
基金
农业部资源遥感与数字农业重点开放实验室开放课题(RDA0808)
国家自然科学基金项目(30800126)
浙江省农业科学院重点实验室资助项目
浙江省农业科学院博士启动项目
关键词
水稻氮素
遥感
统计模型
神经网络
CBERS-1影像
nitrogen content of rice
remote sensing
regression model
neural network
CBERS-1 imagine
