烤烟氮、碱量的冠层光谱检测

Monitoring for total nitrogen and nicotine contents of flue-cured tobacco based on canopy reflectance spectrum

为了实现烤烟氮、碱量田间动态变化的实时无损监测,本试验设计了基于MSR216型多光谱辐射计的烤烟氮、碱量监测试验,在不同烤烟品种的基础上设置了3个氮素水平,用多光谱辐射计采集各处理烤烟的冠层光谱数据,选用了主成分分析法及人工神经网络方法对烤烟6个生长时期的氮、碱量进行了估算,并对2种方法的估算结果进行检验分析。结果显示:冠层光谱检测结果主成分分析法估算的氮、碱量决定性系数分别为0.68、0.58,均方根差分别为0.70%、0.65%;人工神经网络估算的氮、碱量决定性系数分别为0.94、0.92,均主根差分别为0.30%、0.26%,比较而言,人工神经网络估算效果最好,主成分分析法次之。综上而言,利用冠层光谱能够较好地监测和跟踪烤烟冠层氮碱量的动态变化。

The real-time dynamic changes of total nitrogen and nicotine contents of flue-cured tobacco varieties was non-invasively monitored by MSR216 multi-spectral radiometer. The principal component analysis （PCA） and artificial neural network （ANN） were adopted to estimate the total nitrogen and nicotine contents at six growth periods of flue-cured tobacco. The determination coefficients （R2 ） of nitrogen content and nicotine content of the validated models by PCA were 0. 68 and 0. 58, respectively, and the root-mean-square errors （RMSE） were 0. 70% and 0. 65%, respectively. The determination coefficients （R2 ） of the validated models by ANN were 0. 94 and 0. 92, respectively, and RMSE were 0. 30% and 0. 26%, respectively. By comparison, the estimated effect of ANN was better than PCA.