The Novel Combination of Nano Vector Network Analyzer and Machine Learning for Fruit Identification and Ripeness...

Area of Science:

• Agricultural Technology
• Machine Learning Applications
• Spectroscopy

Background:

• Fruit classification is crucial for smart farming and retail.
• Existing methods like image processing and near-infrared spectroscopy (NIRS) have limitations.
• Accurate identification of fruit species and ripeness is needed for various applications.

Purpose of the Study:

• To propose a fast, cost-effective fruit classification method.
• To evaluate the performance of machine learning models (KNN and Neural Network) using VNA data.
• To assess the system's capability for both fruit recognition and ripeness level determination.

Main Methods:

• Utilized a low-cost Vector Network Analyzer (VNA) device.
• Extracted S-parameters (S11 and S21) as features representing signal amplitude and phase.
• Applied K-nearest neighbor (KNN) and Neural Network (NN) models for classification.
• Tested the approach on datasets of five fruit types: Apple, Avocado, Dragon Fruit, Guava, and Mango.

Main Results:

• The Neural Network model achieved high classification accuracy (98.75% and 99.75%) on the first dataset.
• The KNN model demonstrated superior performance in classifying fruit ripeness (98.4%) compared to the NN (96.6%).
• The VNA-based approach proved effective for both fruit recognition and ripeness assessment.

Conclusions:

• A VNA-based system combined with machine learning offers a viable and efficient solution for fruit classification.
• The proposed method is fast, cost-effective, and suitable for smart farming and industrial applications.
• Different machine learning models show varying strengths in fruit recognition versus ripeness classification.