基于模糊神经网络的水稻农田重金属污染水平高光谱预测模型

以吉林省长春一汽厂附近3块不同重金属污染状况的水稻实验样地为例,通过地面实测获取了水稻的光谱反射率、叶片叶绿素含量、叶片及土壤重金属含量等数据.同时,在分析重金属对水稻叶片叶绿素含量影响的基础上,通过多元逐步回归分析选出对水稻叶片叶绿素含量微小变化指示灵敏的光谱参数作为模型输入层,并将水稻叶片叶绿素含量值作为输出层来表征农田重金属污染胁迫水平,最终建立了用于预测水稻农田重金属污染水平的模糊神经网络模型.结果表明,该模糊神经网络模型预测的水稻重金属污染胁迫水平与实测结果吻合度较高,预测的叶绿素含量值与实测值的拟合度较好(R2=0.985).表明在受重金属污染胁迫的情况下,水稻叶片叶绿素含量微小而复杂的变化可以通过构建模糊神经网络模型很好地模拟出来,从而确定出农田的重金属污染水平.

Fuzzy neural network model for predicting stress levels in rice fields polluted with heavy metals using hyperspectral data

Spectral reflectance of rice, chlorophyll content, leaf and soil heavy metal content were collected from three experimental rice fields with different heavy metal pollution levels in Changchun city, Jilin province, China. Based on the analysis of the effect heavy metals on chlorophyll in rice, some spectral indices which are sensitive to subtle changes of the chlorophyll were selected as the input parameters of model with multiple stepwise regression method, and the chlorophyll concentration was used as the output parameter to characterize the stress levels of heavy metal pollution. Finally, the fuzzy neural network model was established to predict the heavy metal pollution levels of rice fields. The results showed that predicated pollution levels corresponded well to measured pollution levels. The correlation coefficient (R2) between measured and predicated chlorophyll concentration was 0.985. It indicated that fuzzy neural network model can precisely simulate the subtle changes of rice chlorophyll content under heavy metal pollution and estimate the heavy metal pollution levels in rice field.