Design of self-adaptable autonomous robotic sprayer for efficient management of tomato leaf diseases

Abstract

This work presents the integration of a self-adaptable precision sprayer with advanced deeplearning approaches for disease detection in agricultural robots.The design features a dynamic scissor mechanism that allows for precise adjustments in both width and height.This enables the equipment to easily adapt to different crop geometries, particularly in the case of tomatoes, wheat, sugar, and potatoes.The ability to adapt is crucial in order to effectively tackle the diverse array of disease threats that these crops face.An integrated deep-learning model has been implemented to address the pressing issue of disease identification, prompted by the elevated disease prevalence in crops.The model improves disease prediction accuracy by utilizing a rich dataset of tomato leaf pictures and employing advanced techniques such as Convolutional Neural Networks (CNN), VGG-16, VGG16 with transfer learning, ResNet with Transfer Learning, and VGG with XGBoost.The system uses smart sensing to detect pests and weeds and distribute chemicals selectively.This reduces waste and environmental effects and improves disease prediction, which is difficult to do manually in fields.The model predicts nearly ten tomato disease classes, with VGG-16 being the most accurate at 94.70%.This research shows how the sprayer's mechanical design and deep learning model work together, advancing precision agriculture technologies.The sprayer's mechanical adaptability and AI-driven disease detection create a flexible, sustainable, and efficient answer for modern farming problems.