基于线性回归模型的菜花重量预测研究

Research on weight forecast of cauliflower based on linear regression model

针对传统测量重量的方法不适于产前作物产量预测的问题,提出1种通过线性回归模型利用菜花三维模型特征属性得到菜花重量的方法。本研究利用Kinect获得菜花三维模型,再通过获取的菜花三维模型的长、宽、高、最大横截面积和体积等属性建立重量预测模型。引用相对误差、决策系数R2作为模型预测精度的评价指标,对线性回归模型正则化得到的岭回归、LASSO模型进行比较分析。结果表明,正则化后模型的泛化能力更强,LASSO模型对菜花重量的预测精度更高。

Regarding the problem that the traditional method of weight measurement is not suitable for prenatal production forecast of crops,we proposed a way of obtaining the weight of cauliflower by using linear regression model and the 3 D model characteristics of cauliflower.The 3 Dmodel of cauliflower is obtained by using Kinect,and the weight prediction model is then established by acquiring the features of the 3 Dmodel like length,width,height,maximum cross section area and volume.The relative error and coefficient of deter mination are used as the evaluation index of the model for accuracy prediction,and the ridge regression and LASSO obtained by regularization are compared and analyzed.The results show that the generalization ability of the regularized models is stronger and the LASSO’s prediction accuracy of cauliflower weight is higher.