<p>Efficient classification and sorting of fruits and vegetables play a vital role in advancing agricultural automation, directly impacting quality control and production efficiency. This study focuses on improving the precision and effectiveness of classifying agricultural produce by utilizing advanced image processing techniques integrated with powerful machine learning algorithms. The research explores several classification models, including Support Vector Machines (SVM), Random Forest, AdaBoost, XGBoost, Decision Tree, Naïve Bayes, and a fine-tuned Convolutional Neural Network (CNN). The proposed framework introduced a novel multi-descriptor feature fusion strategy that combines Histogram of Oriented Gradients (HOG), Local Binary Patterns (LBP), and color-based features to capture both low-level texture and high-level visual information, offering an enriched representation of agricultural images. Additionally, data augmentation is applied to improve robustness and mitigate overfitting. Experimental evaluations were conducted using a publicly available Kaggle dataset comprising over 25,000 labeled fruit and vegetable images spanning multiple classes, lighting conditions, and orientations to ensure diversity. The system achieved a classification accuracy of 99.0%, and F1-score of 99% primarily attributed to the synergistic combination of the proposed feature fusion approach and fine-tuned CNN architecture. To validate generalization, cross-validation techniques were employed, confirming consistent performance across diverse test sets. These results highlight not only the methodological novelty of the integrated fusion–CNN framework but also its strong potential for real-world deployment in automated sorting and quality assurance systems. The proposed model thus offers a scalable and generalizable solution that advances the state of the art in intelligent agricultural automation. </p>