基于叶绿素荧光图像的辣椒叶片氮含量的预测

提取辣椒叶片的25个叶绿素荧光图像的特征参数,其中18个特征参数与氮含量呈极显著相关(P0.01)。用主成分分析法(PCA)提取主要特征参数,将其结果作为遗传算法优化的反向传播人工神经网络(BPNN)、广义回归神经网络(GRNN)和多元线性回归(MLR)模型的输入变量,分别建立辣椒叶片氮含量的预测模型,建模集的相关系数分别为0.959 2、0.963 3、0.943 5,预测集的相关系数分别为0.914 5、0.821 3、0.774 1。

Detection of pepper leaves nitrogen content in greenhouse based on chlorophyll fluorescence image

25 feature parameters were extracted from chlorophyll fluorescence image of pepper leaf, including 18 parameters which was significantly correlated with the nitrogen content at the 0.01 level. Principal component analysis (PCA) was used to extract the main parameters as input variables of genetic algorithm to optimize back–propagation artificial neural network (BPNN), generalized regression neural network (GRNN) and multiple linear regression (MLR), to establish the forecast model of hot pepper leaf nitrogen content, respectively. The correlation coefficient of three model set were 0.959 2, 0.963 3, 0.943 5, and correlation coefficient of prediction set were 0.914 5, 0.821 3, 0.774 1, respectively.