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.pdf

30 10 Vo.l 30, No. 10 2010 10 Acta Scientiae Circumstantiae O ct. , 2010 , , . 2010. [ J]. , 30( 10) : 210 - 2115 L iM, Liu X N, L iu M L. 2010. Fuzzy neuralnetw orkm odel for predicting stress levels in rice fields po lluted w ith heavy m etals using hyperspectral data [ J]. Acta Scientiae Circum stantiae, 30 ( 10) : 210 - 2115 * 李蜜, 刘湘南 , 刘美玲 中国地质大学信息工程学院, 北京 1000 3 : : : 2010062 : 3, . , , , , . , , (R2 = 0. 9 5) . , , . : ; ; ; : 0253246 ( 2010) 10210 0 : X53 : A Fuzzy neural network model for predicting stress levels in rice fields polluted w ith heavym etals using hyperspectral data * LIM ,i LIU X iangnan , LIU M eiling School of Information Engineering, China U niversity of Geosciences, Beijing 1000 3 R eceived 25 January 2010; received in revised form 23M ay 2010; accepted 2 June 2010 A bstract: Data for the spectral reflectance of rice, chlorophyll content, leaf and soil heavy m etal contentw ere collected from three expermi ental rice fields w ith different heavy m etalpo llution levels inChangchun city, Jilin prov ince, China. Based on analysis of the effect of heavy m etals on chlorophyll in rice, spectra l indicesw hich are sensitive to subtle changes of the chlorophyll contentw ere selected as the input param eters of a m odel using multiple stepw ise regression, and the chlorophyll concentration w as used as the output parameter to characterize the stress levels of heavy m etal pollution. F inally, a fuzzy neural netw ork m odelw as established to predict the heavy m etal pollution levels of rice fields. The predicted pollution levels corresponded w ell to the m easured pollution levels. The corre