<![CDATA[Post-harvest quality assessment remains a major challenge in agriculture, particularly for okra (Abelmoschus esculentus), which deteriorates rapidly due to high moisture content. Traditional grading based on manual inspection often results in inconsistency and product damage. This study explores thermal imaging as a non-destructive alternative for okra maturity classification. A dataset of 501 thermal images was acquired under controlled conditions and analyzed using three machine learning models: Convolutional Neural Network (CNN), Support Vector Machine (SVM) with Histogram of Oriented Gradients (HOG) features, and Long Short-Term Memory (LSTM) network. Experimental results show that CNN achieved the highest accuracy (99.01%), outperforming SVM (95.05%) and LSTM (91.09%). Confusion matrix and ROC analyses confirmed CNN’s superiority in capturing spatial thermal patterns related to maturity stages. Compared with RGB or hyperspectral imaging reported in prior studies, thermal imaging integrated with AI provides a more robust, illumination-independent, and non-destructive solution. The findings demonstrate the potential of CNN-based thermal imaging systems for automated sorting of okra in agricultural supply chains. Future work will focus on larger datasets, multi-class maturity levels, and real-time implementation to enhance practical deployment in post-harvest management.]]>
Copyrights © 2025
