Design, Fabrication, and Evaluation of a Non-Destructive Testing System for Improving Durian Ripeness Testing

With its economic and cultural value, durian plays a significant role in Southeast Asia. However, its ripeness is challenging due to the thick, spiky shell that conceals its edible interior. Traditional tapping methods remain widely used but are subjective and often unreliable. The system combines an InfiRay P2 Pro infrared camera with an EfficientNet-B0 convolutional neural network that was only trained on thermal image features. The system was tested on 1,000 durian samples, showing that 87.5% were correctly classified. The proposed system shows strong potential for smart agriculture by enabling reliable, real-time ripeness classification to assist farmers and future automated grading solutions.