生菜叶片含水率光谱特征模型研究.PDFVIP

20 1 1年5 月 农 业 机 械 学 报 第42 卷 第5 期 生菜叶片含水率光谱特征模型研究* 毛罕平摇 高洪燕摇 张晓东 (江苏大学现代农业装备与技术省部共建教育部重点实验室,镇江212013) 摇 摇 揖摘要铱摇 利用便携式光谱分析仪测量生菜叶片的光谱反射率,并对其进行对数变换。 通过变量筛选得到 725、1075、1272、1450、1640 和1958nm波长处的光谱反射率与生菜干基含水率呈极显著相关。 为克服多重共线 性影响,分别采用多元线性回归分析、主成分回归分析、偏最小二乘回归分析及偏最小二乘 人工神经网络回归分 析4种方法建立了叶片干基含水率的定量分析模型。 结果表明4种算法预测值与实测值相关系数分别为0郾4850、 0郾8992、0郾9174 和0郾9470,偏最小二乘 人工神经网络模型的预测能力优于其他模型。 关键词:水分胁迫摇 生菜摇 叶片摇 含水率摇 光谱特征摇 数学模型 中图分类号:O657郾33;Q945郾17 文献标识码:A 文章编号:1000鄄1298(2011)05鄄0166鄄05 Spectral Characteristics Model of Lettuce Leaves蒺Water Content Mao Hanping摇 Gao Hongyan摇 Zhang Xiaodong (Key Laboratory of ModernAgricultural Equipment and Technology,Ministry of Education JiangsuProvince, Jiangsu University,Zhenjiang 212013,China) Abstract Spectral reflectanceof lettuceleavesingrowing statuswasmeasuredusingtheASDFieldSpec3,and logarithmic transformation was also obtained. By variable selection, it was found that the linear relationships betweendry鄄basismoisturecontentoflettuceleavesandspectralreflectancedatain725nm, 1075nm,1272nm,1450 nm,1640 nm and 1958 nm were very notable. In order to overcome the impact of multicollinearity, quantitative analysis models of dry鄄blade爷s moisture content have been established respectively with methods of multiple linear regression analysis, principal component regression analysis,partial least squares regression analysis and PLS鄄neural network analysis. The result showed that the correlation coefficient R of measured and predictive valuesfrom the four algorithms were 0郾4850,0郾8992,0郾9174 and0郾9470 respectively,which showedbetter predictiveperformance of the model based on PLS鄄neural network a