基于GA-ICA和高光谱图像技术的黄瓜叶叶绿素检测

摘要

用遗传算法(GA)和独立分量法(ICA)提取黄瓜叶高光谱图像的特征信息,对高光谱图像技术检测叶绿素含量及其叶面分布的可行性进行了研究.高光谱图像标定校正后,提取其中的光谱信息,采用GA对光谱信息进行特征波长选择,将GA优选出来的光谱进行ICA信号分析并结合叶绿素含量值建立多元线性回归模型(MLR).结果表明GA共优选出280个特征波长,在此基础上提取光谱的8个ICA信号建立叶绿素含量MLR模型,模型对应的预测集相关系数为0.931 2,对应的预测均方根误差为0.191 4.提取所有像素点光谱在特征波长下对应的8个ICA信号,代入建立的叶绿素含量模型中,快速计算出所有像素点对应的叶绿素含量,得到黄瓜叶叶绿素含量叶面分布图.研究结果表明:利用高光谱图像技术结合GA与ICA快速、无损检测叶片叶绿素含量及其叶面分布是可行的.%Chlorophyll distribution in cucumber leaves was non-destructively and rapidly measured based on genetic algorithms (GA) combined with independent component analysis (ICA) by Hyper-spectral images technique.Spectrum was extracted from hyper-spectral images of cucumber leaves after pre-processing.Firstly GA was used to search for an optimum informative wavelength of chlorophyll, then ICA signals were computed to build multiple regression model (MLR).The results show that 280 wavelengths were selected by GA and MLR model which was obtained based on 8 ICA signals.MLR model performs well with prediction coefficients of 0.931 2 and root mean standard error of prediction (RMSEP) of 0.191 4.8 ICA signals of every pixel in hyper-spectral images were computed and chlorophyll content of every pixel was obtained according to the MLR model.Finally, the chlorophyll distribution map was estimated.Overall results sufficiently demonstrate that the hyper-spectral imaging technique can be used to measure the chlorophyll concentration and to estimate the distribution of chlorophyll in cucumber leaf.