Towards A Decision Support System for Farm Precision

This research addresses challenges in agriculture arising from climate change and unpredictable environments, which significantly impact crop production. We propose a framework to tackle these challenges using machine learning for pest identification. The framework consists of three parts: i) Data Collection: we utilize Arduino and TensorFlow to capture images of insects in the field. TensorFlow processes the images using a trained machine learning algorithm to identify insect characteristics like type, species, and location. Additionally, TensorFlow is used to generate new pest images to enrich the training dataset. ii) Machine learning model development: a random forest model is used to identify existing insects. The model's performance is evaluated using precision, recall, and accuracy metrics. Results show high accuracy (93.33%) across different insect classes. And iii) web application development: the trained model is integrated into a web application to identify pests from user-uploaded images. This information can be used to automate a pesticide spraying system for targeted pest control. Additionally, this research contributes to precision agriculture by offering a framework for pest identification and supporting the development of automated spraying systems, potentially reducing pesticide overuse and environmental impact.