<![CDATA[General Background: Rising food demand and climate variability require precise, scalable crop monitoring solutions. Specific Background: Traditional field inspections are labor-intensive, subjective, and unsuitable for large areas, motivating image-driven automation. Knowledge Gap: Many studies address plant disease detection, yet few present an integrated, adaptable framework that unifies preprocessing, feature learning, and multi-class crop condition assessment under diverse field conditions. Aims: This study develops a machine learning image analysis system using convolutional neural networks to classify crops as healthy, normal, or diseased from ground, UAV, and remote-sensing images. Results: The model achieved stable, high-accuracy classification, strong recall for diseased crops, and robustness to lighting, background variability, and crop diversity through preprocessing and augmentation. Novelty: The work integrates end-to-end preprocessing, deep feature extraction, and comparative positioning against SVM and KNN within a unified monitoring pipeline tailored to real-field variability. Implications: The system supports timely agro-technical decisions, reduces human error, and advances practical smart farming and digital agriculture deployment. Highlights: End-to-end CNN pipeline for healthy, normal, and diseased crop classification. Robust performance under variable lighting, background, and crop types. Practical pathway toward scalable smart farming monitoring systems. Keywords: Crop Monitoring, Convolutional Neural Networks, Image Processing, Smart Farming, Machine Learning]]>
