An Indirect Method to Estimate Sweet Lime Weight through Machine Learning Algorithm

Abstract

A fast and indirect method of weighing the sweet lime fruit developed based on the computer vision coupled with machine learning algorithm is investigated in this research work. The developed computer vision system (CVS) has been used to analyze the sweet lime image database. The images have been processed using the developed algorithm to extract seven geometrical attributes. The support vector machine regression (SVMR) modelling technique has been utilized to develop the model for estimating the weight of fruit samples under consideration. Eight different SVMR models have been developed in two SVM type for different kernel type. Relevant statistical analysis and comparison of the developed model is also presented. Finally, the type 2 SVMR model with RBF kernel has been recommended as the model with best performance during training ($R^{2}=$ 0.9867, RMSE = 5.26) and testing ($R^{2} =$ 0.9866, RMSE = 6.435) too. Thus, the presented work provides an indirect way for measuring sweet lime fruit size to estimate its weight. This will be helpful in the design and development of most of the post-harvest equipment.