AI for Crop Disease and Pest Detection Using Remote Sensing and Computer Vision: An Empirical Study

Abstract This research investigates deployment of AI-based models like ResNet50 to detect crop diseases/pests helped with Remote Sensing and Computer Vision techniques. As agriculture becomes more exacting and efficient, a method for detecting diseases and pests early would help reduce crop losses and pesticide application. The process of training the ResNet50 model with labelled images over 20000 plus crop images and following the data pre-processing, feature extraction and evaluating the model. Evaluation of performance metrics including accuracy (91.3), precision (90.1), recall (92.5), and F1 score (91.3) show efficiency of the model in disease classification. The results demonstrate that ResNet50 surpasses others such as VGG16, SVM, and Random Forest, with Disease 2 demonstrating the greatest detection accuracy of 98.5%. The confusion matrix revealed low misclassification rates, especially for healthy crops and Disease 2. However, Disease 1 had a relatively higher false discovery rate, which can be improved upon. The model accuracy was evaluated based on cross-validation results across five-folds achieving a mean accuracy of 91.3%. Using an AI-based model such as ResNet50 will give people high accuracy detection which can help a lot to raise precision disease management in agriculture. Future work should focus on extending the datasets, developing the model’s architecture and deploying real-time detection in the field. AI-based pest and disease detection systems should be integrated into precision farming for improving crop yield, reducing the use of pesticides, and encouraging sustainable farming practices.