Application of advanced Artificial Intelligence models to manage irrigation using sensor data and satellite images

Résumé

In this thesis, we tackle the urgent challenges in water management, agricultural image analysis, by the investigation of artificial intelligence (AI) models to optimize precision farming. Recognizing the growing threat of climate change and the global water crisis, we conducted a systematic review of smart irrigation technologies, focusing on IoT sensors, remote sensing, and AI methods. This comprehensive review not only highlights existing approaches but also sets the stage for new solutions that optimize water use and enhance agricultural sustainability. To address the critical lack of agricultural datasets, we proposed and implemented convolutional neural networks (CNNs) and generative adversarial networks (GANs) to predict soil moisture from UAV-captured aerial images. By proposing a novel GAN model that generates conjointly synthetic images and their continuous ground truth vectors, we significantly enhanced CNN performance in data-scarce environments. This significantly reduced prediction errors, proving the power of GAN-driven data augmentation in regression tasks, a data augmentation setting not handled by conventional GANs. Additionally, we have proposed Hybrid AI models, combining deep learning models with Machine learning models leveraging human expert-based features, for predicting nitrogen content in sorghum crops—an essential factor for crop health—using UAV-captured RGB imagery. We have carried out this research in the context of collaboration with the Czech University of Life Sciences in Prague (CZU) and the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT) in Telangana, India. By integrating spectral indices with CNN architectures, we enhanced the accuracy of nitrogen predictions, supporting more precise and sustainable agricultural practices. Through the fusion of IoT, AI, and RS technologies, our work provides innovative solutions to address critical challenges in water resource management and environmental sustainability.

Citer ce document

Accès au document

Ce document est hébergé sur son dépôt institutionnel d'origine.

Auteur(s)

Statistiques

Consultations : 2

Téléchargements : 0