Unmixing-Based Fusion of Hyperspatial and Hyperspectral Airborne Imagery for Early Detection of Vegetation Stress
Derechos de accesoopenAccess
MetadatosMostrar el registro completo del ítem
AutorDelalieux,Stephanie; Zarco-Tejada,Pablo J.; Tits,Laurent; Jimenez Bello,Miguel Angel; Intrigliolo,Diego S.; Somers,Ben
Many applications require a timely acquisition of high spatial and spectral resolution remote sensing data. This is often not achievable since spaceborne remote sensing instruments face a tradeoff between spatial and spectral resolution, while airborne sensors mounted on a manned aircraft are too expensive to acquire a high temporal resolution. This gap between information needs and data availability inspires research on using Remotely Piloted Aircraft Systems (RPAS) to capture the desired high spectral and spatial information, furthermore providing temporal flexibility. Present hyperspectral imagers on board lightweight RPAS are still rare, due to the operational complexity, sensor weight, and instability. This paper looks into the use of a hyperspectral-hyperspatial fusion technique for an improved biophysical parameter retrieval and physiological assessment in agricultural crops. First, a biophysical parameter extraction study is performed on a simulated citrus orchard. Subsequently, the unmixing-based fusion is applied on a real test case in commercial citrus orchards with discontinuous canopies, in which a more efficient and accurate estimation of water stress is achieved by fusing thermal hyperspatial and hyperspectral (APEX) imagery. Narrowband reflectance indices that have proven their effectiveness as previsual indicators of water stress, such as the Photochemical Reflectance Index (PRI), show a significant increase in tree water-stress detection when applied on the fused dataset compared to the original hyperspectral APEX dataset (R-2 = 0.62, p 0.1). Maximal R-2 values of 0.93 and 0.86 are obtained by a linear relationship between the vegetation index and the resp., water and chlorophyll, parameter content maps.