基于反射光谱的苹果叶片叶绿素和含水率预测模型  

Forecasting Chlorophyll Content and Moisture of Apple Leaves in Different Tree Growth Period Based on Spectral Reflectance

作　　者：冀荣华[1] 郑立华[1] 邓小蕾[1] 张瑶[1] 李民赞[1] 

机构地区：[1]中国农业大学信息与电气工程学院,北京100083

出　　处：《农业机械学报》2014年第8期269-275,共7页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家自然科学基金资助项目(31071330)

摘　　要：为探索苹果叶片叶绿素含量（质量比）、叶片含水率与反射光谱之间的关系,以华北地区苹果树为研究对象,分别测定了各个关键生长期苹果叶片的光谱反射率、叶绿素含量和叶片含水率。分析光谱反射率与叶绿素含量以及叶片含水率之间相关性发现,在不同生长时期,苹果叶片叶绿素a含量与反射光谱在515~590 nm和688~715 nm两组波段内具有较高的相关性,且果实成熟期数据显示相关度最高（R2=0.6）。在420~500 nm、640~680 nm、740~860 nm 3个波段叶片含水率与反射光谱有较高的相关性,且果实膨大期的叶片含水率在可见光波段的相关系数最大。根据所选敏感波段,分别利用多元线性回归、主成分分析和人工神经元网络建立基于反射光谱的苹果叶片不同生长时期叶绿素和含水率的预测模型。通过对所建立的预测模型进行校验,结果显示,利用主成分分析方法所建立的苹果叶片叶绿素含量预测模型的决定系数最高（R2=0.885 2）,校验系数为0.828 9。该模型可以较为准确地预测苹果叶片叶绿素含量。而采用神经元网络所建立苹果叶片含水率预测模型的决定系数R2=0.862,校验系数为0.8375,预测效果最好。In order to detect the growth status of apple trees based on spectroscopy, an apple orchard was selected as the experimental site located at the outskirts of Beijing. First, the samples of apple tree leaves at each key growth period were collected. Then the spectral reflectance, chlorophyll content and moisture content of the samples were measured respectively. The characteristics of those spectra were analyzed and the correlation between chlorophyll content, moisture content and their spectra were calculated. The results showed that the original spectra were most correlated with leaf 590 nm and from 688 nm to 715 nm. The correlation coefficients chlorophyll content from 511 nm to in September were high and the maximum value was 0.6. From the correlation analysis between apples leaves moisture content and their spectra, it was found that the original spectra were most correlated with leaf moisture content at the wavebands of 420 ~ 500 nm, 640 ~ 680 nm and 740 ~ 860 nm, and the correlation coefficients in fruiting period were high. According to the selected sensitive bands, the models for estimating the chlorophyll content and moisture content in apple leaves were built by multiple linear regression analysis （MLRA） ,principal component analysis （PCA） and artificial neural network （ANN）, respectively. The models were tested by the validation set which included 25 samples of apple tree leaves. The forecasting results indicated that the model based on PCA was the best model to predict the chlorophyll content of apple leaves, and the calibration and validation R2 were 0. 885 2 and 0. 828 9, respectively. The forecasting model of apple leaf moisture content based on ANN was the best, and the calibration and validation R2 were 0. 862 and O. 837 5, respectively.

关 键 词：苹果叶片 光谱 叶绿素含量 含水率 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]