Proceedings
Authors
| Filter results4 paper(s) found. |
|---|
1. Weed mapping: advantages of RGB CNN-based approaches vs multispectral pixel-based methodsWeed detection remains a major challenge in modern agriculture, and accurate weed mapping is crucial to support rapid and efficient management interventions, ensuring crop productivity and economic viability. In this context, geotechnologies such as remote sensing and computer vision, together with the widespread adoption of drones, enable the acquisition of ultra–high spatial resolution imagery, allowing more detailed analyses in complex agricultural environments. Although multispectral... |
2. Integration of LiDAR-Derived TWI and UAV Multispectral Data for Waterlogging Susceptibility Mapping in Precision AgricultureTopography directly controls water redistribution across the landscape, shaping the spatial variability of soil moisture in agricultural areas. The Topographic Wetness Index (TWI), derived from digital elevation models, is widely used to estimate the potential for water accumulation; however, its field-scale validation supported by high-resolution multispectral drone imagery remains limited. In agricultural systems, recurrent waterlogging can reduce productivity by impairing germination, promoting... |
3. Weed identification in soybean fields using RGB UAV imagery acquired at different flight altitudesThe presence of weeds in agricultural fields is one of the main factors reducing crop productivity due to competition for light, water, and nutrients. In this context, digital agriculture and the use of unmanned aerial vehicles (UAVs) enable the acquisition of high-resolution imagery for detecting and monitoring these weeds. However, increasing flight altitude reduces spatial resolution, compromising the identification of key visual attributes (shape, texture, and edges) and making it more difficult... |
4. Generation of Ultra-High-Resolution Synthetic Data via Generative Super-Resolution to Support UAV Image Annotation and Model TrainingManual annotation of imagery acquired by unmanned aerial vehicles (UAVs) for detection/segmentation tasks is one of the main bottlenecks for deep learning applications in precision agriculture, due to the high cost and the time required to produce consistent labels. In addition, low-altitude flights to obtain ultra–high spatial resolution increase operational complexity and data volume, limiting the scalability of acquisition campaigns. Although neural network–based super-resolution... M.A. Karasinski, R. Costa, C. Melville, E. Macedo, I.L. Gabriel Da Silva Carmo , S.V. Dantas Oliveira, M.P. Galvão, A.B. Bendahan, C.R. Bezerra |