Paper 13197-33
Paper 13197-33
A comparative analysis of machine learning algorithms for maize crop mapping
18 September 2024 • 11:10 - 11:30 BST | Moorfoot
Abstract
Accurate mapping of maize crops is crucial for sustainable agriculture and food security in developing countries. This study explores the application of machine learning algorithms, specifically Random Forest (RF) and k-Nearest Neighbors (kNN), to map maize crops in South Africa's Free State province. Using optical Sentinel-2 and Sentinel-1 radar images, the RF algorithm achieves an impressive 90% accuracy, while the kNN algorithm demonstrates a notable 86% precision. These findings provide valuable information for decision-makers and agronomists in selecting optimal mapping approaches. Future studies can build on these insights to refine strategies to address food insecurity and advance sustainable agriculture.
Presenter
Simbarashe Jombo
Sol Plaatje Univ. (South Africa)
Simbarashe Jombo, a postdoctoral research fellow at Sol Plaatje University, specialises in remote sensing, holding a PhD in Geography from the University of the Witwatersrand. His expertise includes vegetation and crop mapping, change detection analysis, and applying machine learning to geospatial analysis. His recent focus involves the development of remote sensing methods, particularly using machine learning for urban tree species classification with LiDAR, hyperspectral, and multispectral data. Currently, Dr. Jombo is actively involved in a project addressing arid rangeland degradation in South Africa through remote sensing applications. This research has significant implications for sustainable maize production in arid regions, contributing valuable insights into the monitoring and modeling of agricultural landscapes. Dr. Jombo's work aligns with efforts to enhance food security by leveraging advanced technologies and methodologies tailored to the challenges of arid environments.