基于图像处理的水培生菜冠层图像叶面积估测研究

Leaf area estimation of hydroponic lettuce canopy based on image processing

为实现精准、高效、无损地获取植物工厂环境下水培生菜相关长势参数叶面积(Leaf area,LA),基于数字图像处理和机器学习回归方法建立单株水培生菜冠层图像LA估测模型。首先,通过智能手机获取2个生菜品种不同生长期的冠层可见光图像,利用Photoshop图像处理软件将原始图像统一剪裁为900像素×900像素大小,采用中值滤波(MedianBlur)法对剪裁后的图像进行去噪运算,将RGB图像转化为HSV颜色空间,再采用mask掩膜法分割彩色图像;然后,利用图像法获取单株生菜LA实测值,构建以LA实测值为因变量,以生菜冠层投影面积(Projected leaf area,PLA)为自变量的线性回归(Linear regression,LR)模型和以全局图像特征(颜色、形状、纹理等)为自变量的支持向量回归(Support vector regression,SVR)、多元线性回归(Multiple linear regression,MLR)和随机森林(Random forest,RF)等LA估测模型进行对比分析;最后,采用决定系数(Coefficient of determination,R^(2))和均方根误差(Root mean square error,RMSE)评估模型的准确性。结果表明:RF模型估测效果最好,对于生菜品种‘绿萝’单株LA估测结果的R^(2)为0.9714、RMSE为8.89 cm2,对于品种‘碧霄’估测结果的R^(2)为0.9201、RMSE为23.34 cm2。本研究验证了RF回归模型能够较准确地估测生菜单株叶面积,可为植物工厂水培生菜LA无损估测提供新的解决方案和研究基础。

To achieve accurate,efficient,and non-destructive acquisition of leaf area(LA)parameters related to hydroponic lettuce growth in plant factory environments,single hydroponic lettuce canopy image LA estimation models were established based on digital image processing and machine learning regression methods.Firstly,visible light images of the canopy of 2 lettuce varieties at different growth stages were obtained through a smartphone.The original images were uniformly cropped to a size of 900 pixels×900 pixels using Photoshop image processing software.The MedianBlur method was used to denoise on the cropped images,converting the RGB images into HSV color space,and the color images were segmented using the mask method.Then,using the image method to obtain the measured LA values of individual lettuce plant,LA estimation models such as linear regression(LR)with LA measured values as the dependent variable and the projected leaf area(PLA)of lettuce canopy as the independent variable,and support vector regression(SVR),multiple linear regression(MLR),and random forest(RF)with global image features(color,shape,texture,etc.)as independent variables were constructed for comparative analysis.Finally,the accuracy of the models were evaluated using the coefficient of determination(R^(2))and root mean square error(RMSE).The results showed that the RF model had the best estimation effect,with R^(2)of 0.9714 and RMSE of 8.89 cm2 for the single plant LA estimation of lettuce‘Lvluo’,and R^(2)of 0.9201 and RMSE of 23.34 cm2 for‘Bixiao’.This study verified that the RF regression model could accurately estimate the leaf area of lettuce plants,providing a new solution and research basis for non-destructive estimation of hydroponic lettuce LA in plant factories.