CNN-Based Automated Detection of Wax Coating on Apples

Implementing precise and effective quality control measures during postharvest processing is essential for upholding the standards of agricultural products. Conventional approaches to evaluating wax coatings on apples, commonly utilized to improve glossiness and extend shelf life, frequently depend on laborintensive and subjective manual evaluations, resulting in inconsistencies and inefficiencies. This study introduces a CNN-based model aimed at automated detecting wax coatings on apples, thereby providing a robust and scalable solution to address existing gaps in the field. Utilizing a diverse dataset comprising high-resolution apple images has resulted in notable improvements, achieving an accuracy of 96.4% in the model's performance. Furthermore, the model demonstrates exceptional performance across critical metrics, achieving a precision of 95.9%, a recall of 96.5%, and an F1-score of 96.2%. Additionally, it results in an AUC-ROC of 0.98, highlighting its substantial discriminative capability. The multiclass classification approach effectively differentiates between the Uncoated, Thin Wax Coating, and Thick Wax Coating categories. However, some minor misclassifications are observed between the thin and thick coatings. The results show that the model can change post-harvest processing by reducing the need for manual work, making things more consistent, and making it easier to check quality in real-time. The proposed approach represents a notable advancement in modernizing agricultural quality control practices, enhancing efficiency and reliability throughout the supply chain.