Thermography and machine learning techniques for tomato freshness prediction.

Area of Science:

• Agricultural Science
• Computer Science
• Food Science

Background:

• Tomato quality and freshness are critical for the multi-billion dollar global industry.
• Existing non-destructive methods for assessing tomato freshness have limitations.
• Accurate prediction of tomato decay remains a significant challenge.

Purpose of the Study:

• To explore thermography's potential for revealing internal tomato states.
• To investigate machine learning (ANNs and SVMs) with transient heating and thermography for freshness prediction.
• To compare the accuracy of ANNs and SVMs in predicting tomato decay.

Main Methods:

• Utilized transient step heating and infrared thermal imaging.
• Captured thermal images at 1 Hz during heating (40s) and cooling (160s) phases.
• Applied artificial neural networks (ANNs) and support vector machines (SVMs) using temperature differences as input features.

Main Results:

• Developed an optimal ANN model with a regression coefficient of 0.99.
• Identified temperature differences in specific regions as key indicators of freshness.
• The ANN model demonstrated higher accuracy in predicting tomato freshness compared to SVM models.

Conclusions:

• Infrared thermal imaging coupled with ANNs provides a highly accurate, non-destructive method for predicting tomato freshness.
• This approach offers a promising solution to the ongoing challenge of assessing fruit quality and predicting decay.
• The study validates the effectiveness of ANNs over SVMs for this specific application.