Automated fruit quality inspection from IR image processing

This paper presents an automated fruit quality inspection system using infrared (IR) image processing and deep learning in MATLAB. The proposed method classifies fruit ripeness stages—early, advanced, and healthy—using a Convolutional Neural Network (CNN)-based architecture. Input images, initially in RGB format, are converted into simulated IR images to enhance feature extraction, improving classification accuracy. The CNN model automatically extracts essential features, enabling robust and precise categorization of fruit ripeness. The study focuses on three fruits—apple, orange, and banana—and is conducted on a controlled dataset to validate the effectiveness of the approach. The use of IR image simulation enhances the model’s ability to distinguish ripeness levels by capturing thermal variations and internal fruit properties. The proposed system offers a non-invasive, efficient, and scalable solution for automated fruit quality assessment, reducing reliance on manual inspection methods. Experimental results demonstrate the model’s capability to accurately classify ripeness stages, showcasing its potential for real-world agricultural and supply chain applications. The integration of deep learning with IR image processing ensures improved feature representation, making this approach suitable for industrial applications in automated sorting and quality control. Future work will explore extending the model to a broader range of fruits and optimizing CNN parameters for enhanced accuracy

Index Terms— Fruit images Dataset, Deep Learning algorithm, Convolutional Neural Network, and Accuracy.