Machine learning techniques for non-destructive estimation of plum fruit weight.

Area of Science:

• Agricultural Science
• Computer Science
• Data Science

Background:

• Accurate plum fresh weight (FW) estimation is vital for agricultural practices like yield prediction and quality control.
• Traditional FW measurement methods are often destructive, labor-intensive, and time-consuming.
• Developing non-destructive techniques for FW estimation is a significant agricultural challenge.

Purpose of the Study:

• To investigate the efficacy of artificial intelligence (AI) and machine learning (ML) for predicting plum fresh weight (FW) using fruit dimensions.
• To evaluate and compare the performance of various ML models in estimating plum FW.
• To establish an efficient, non-destructive method for plum FW estimation.

Main Methods:

• Fruit images captured via smartphone camera were processed to extract dimensions.
• Machine learning models including Support Vector Regression (SVR), Multivariate Linear Regression (MLR), Multi-Layer Perceptron (MLP), and Decision Tree (DT) were employed.
• The SVR model with a Pearson-VII kernel (PUK) function and a penalty value (c) of 0.1 was selected for detailed analysis.

Main Results:

• The SVR model demonstrated high accuracy in plum FW estimation.
• Achieved an R-squared (R²) value of 0.9369 during training and 0.9267 during testing.
• Reported a root mean squared error (RMSE) of 0.4850 g (training) and 0.4863 g (testing), indicating precise predictions.

Conclusions:

• The developed AI-based method using SVR provides an accurate and non-destructive approach for plum fresh weight estimation.
• This technique offers a significant improvement over traditional, destructive methods.
• The findings are valuable for researchers and practitioners in precision agriculture, enabling efficient yield and quality assessments.