AI & Drone-Driven Solution to Improve Crop Yield via Precise Pest Control

Agriculture is vital for ensuring food security and sustaining economic growth, yet it faces substantial threats from pests and diseases, which account for up to 40% of global crop production losses annually. This study presents an innovative AI and drone-driven solution deployed by GeoMinds Africa to combat agricultural pests, specifically targeting the Thaumatotibia (Cryptophlebia) leucotreta (Meyrick) pest, which affects more than 35 kinds of plants across 37 African countries by leveraging artificial intelligence (AI) and UAV. Over 3 months, high-resolution drone imagery (DJI Mavic 3 Multispectral) was captured in Gaya City (Niger) and processed, generating a dataset of over 5,000 labeled images with the active participation of 50 farmers. After compared with other machine learning models (SVM, and Gradient Boosting), a Random Forest model was selected and trained on the labeled images, with the integrated weather data, historical records of pests, and vegetation indices such as Normalized Difference Red Edge Index (NDRE), to identify early signs of Thaumatotibia leucotreta pest infestation. As a result, the study highlights that the implementation of AI and drone technology in Gaya City improved the management of Thaumatotibia leucotreta on mango crops. The Random Forest model achieved 92% accuracy, outperforming other models. NDRE values effectively identified stress areas, with a 10% increase in detection accuracy when integrated with weather data. Field validation showed 90% farmer satisfaction, leading to a 20% reduction in crop losses and a 15% yield improvement for the 50 farmers.