In recent years, advances in computational technologies and new methods have been made available for plant growth and physiology studies. Due to its practicality and precision, studies with modeling for leaf area prediction have been carried out with several agricultural species, being considered a non-destructive method. Thus, the study aimed to develop non-destructive methods based on regression models, support vector machine (SVM) and random forest for predicting the leaf area of Sesamum indicum cultivars using linear leaf dimensions. A total of 9600 leaves were collected from four S. indicum cultivars, and 2400 leaves of each cultivar were collected. The cultivars used for sampling were BRS Seda, CNPA G2, CNPA G3, and CNPA G4. The measurements of each leaf’s length, width, and leaf area were obtained through scanned images using the ImageJ software. Then, the product between length and width was calculated. Allometric equations were constructed using linear, power, and exponential models. The criteria for choosing the best models were the coefficient of determination (R2), mean absolute error (MAE) and root of the mean square of error (RMSE). The SVM learning algorithm can be used with greater accuracy to predict the leaf area of S. indicum cultivars using leaf dimensions, such as length and width. Despite new technological advances in equipment, the methods proposed in this study, such as SVM and regression models, provided accurate predictions for the leaf area of all sesame cultivars.
扫码关注我们
求助内容:
应助结果提醒方式:
