Obtaining and using aerial thermal images for the study of in-field crop spatial variability

• Autor(a) principal: Acorsi, Matheus Gabriel
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Biblioteca Digital de Teses e Dissertações da USP
• Texto Completo: https://www.teses.usp.br/teses/disponiveis/11/11152/tde-04082021-132031/
• topic: Calibração; Calibration; Câmera termográfica; Conteúdo de água no solo; Estresse hídrico; Plant water stress; RPA; Soil water content; Termografia; Thermography; UAV; Uncooled thermal camera

The use of miniaturized thermal cameras attached to unmanned aerial vehicles (UAV) expanded the applications of remotely sensed temperature, especially in small-scale agriculture, where high spatial and temporal resolutions are necessary. However, deriving accurate temperature readings from these cameras is often challenging due to issues related to unstable accuracy and orthomosaic processing. Overcoming these problems is fundamental to ensure the accuracy needed for applications such as crop water stress monitoring, where small differences in canopy temperature (CT) can reflect an onset water stress. In this study, we tested a low-cost thermal camera in proximal and aerial conditions, focusing on developing a feasible methodology to deliver accurate temperature readings and mitigate issues related to orthomosaic processing. The results include a co-registering method that significantly increased the image alignment performance, producing suitable thermal orthomosaics. In terms of accuracy, calibration models were developed according to the flight altitude and conditions tested, resulting in root mean squared errors (RMSE) lower than 2°C, with best results obtained with orthomosaics produced with blending mode set as average. Using this methodology, we investigated the relationship of UAV-based canopy temperature with soil and plant attributes linked to water status in a rainfed maize field. While the aerial images were taken, a guided sampling took place across the field to determine soil and plant water content. The results demonstrated that the field of study presented high spatial variability in terms of soil water storage (SWS), with a coefficient of variation of 23.3% and values close to the permanent wilting point (PWP), confirming the soil water deficit. This result was reflected on CT values, which ranged from 32.8 to 40.6°C among the sampling locations. Although CT correlated well with most of the soil physical attributes related to water dynamic, the simple linear regression between CT and soil water content variables yielded coefficients of determination (R2) ≤ 0.26, indicating that CT alone is not sufficient to predict soil water status. Nonetheless, when CT was combined with some soil physical attributes in a stepwise multiple linear regression the prediction capacity was significantly increased achieving R2 values ≥ 0.83, demonstrating a potential use for CT associated with pre-existing soil attributes as an in-season tool to assess the spatial variability of soil water content.