DICL: Appliance of Deep Learning Approach to Identify Accurate Rice Varieties by Using Digital Image Classification Logic

Abstract

Rice is one of the most widely consumed staple foods across the globe, and accurate identification of its varieties is crucial for ensuring food quality, authenticity, and market value. Conventional manual inspection is prone to human error, inefficiency, and inconsistency, especially when dealing with morphologically similar varieties. In order to address these issues, this paper presents DICL (Digital Image Classification Logic), a deep learning-based image classification model that can classify rice varieties using high-resolution digital images. The technique includes the retrieval of data on agricultural depositories, research centers and field-based photographs which are then pre-processed using resizing, noise reduction, histogram equalization, background removal and augmentation. Convolutional Neural Networks (CNNs) were used to extract features, which were assisted by transfer learning with ResN et, whereas statistical features like entropy and Gray- Level Co-occurrence Matrix (GLCM) improved the textural representation. An architecture with CNN layers, Capsule Networks, and Vision Transformers was created to be robust in classification as it was trained with categorical cross-entropy loss and optimized through stochastic gradient descent with momentum. The experimental outcomes confirm that the proposed DICL model has an accuracy of 97.6% and a precision of 97.1 % and a recall of 96.5% and FI-score of 0.972, which is higher than the traditional classifiers (SVM, Random Forest, and k-NN) and baseline CNN models. The analysis of the confusion matrix was used to verify the decrease in the misclassification levels, especially in those varieties whose morphological similarities were high. The results point to the possibilities of DICL to real life use at rice mills, regulatory inspection, and smart agriculture. This framework will play a significant role in advancing automated crop quality assessment by offering a scalable and interpretable solution which is accurate.