Cotton growth monitoring combined with coefficient of variation method and machine learning model
In order to obtain the growth information of the key phenological period of cotton more accurately,the cotton planting area is extracted through the cotton mapping index.Secondly,five indexes,including plant height,SPAD value,leaf wet weight,leaf dry weight and leaf area,reflecting cotton growth,are constructed into a comprehensive growth Index,namely Flowering and boll cotton growth index(FBCGI),using the coefficient of variation method.Finally,the optimal characteristic variables are selected and the inverse model of cotton growth is constructed by combining with random forest model.The results showed that:① The overall classification accuracy of cotton reached 81.65％.②Compared with the five single growth indicators,the constructed FBCGI had a higher correlation with vegetation index.③The R2 and RMSE of the cotton growth monitoring model based on the optimal characteristic variables and random forest model in the modeling set and validation set are 0.74,0.07 and 0.51,0.10,respectively.The results can provide important reference for cotton growth monitoring.
万方数据
维普
