Synthetic Imagery of Cotton Crops: Scaling from Leaf to Full Canopy

Advances in sensing technology are creating the rapid development of remote imaging systems for crop monitoring. Sensors are able to distinguish crop stress, such as that associated with low nutrient status, herbicide-induced stress, insect pressure and poor stand establishment. Most of these systems rely on the reflectance properties of natural pigments present in leaf tissue (carotenoids and chlorophylls), and detect the production of alternate compounds under stress conditions (e.g. xanthophylls) to indicate plant stress. Reflectance properties of these plant compounds in the visible region have led to the development of remote imagery systems in the visible and near-infrared regions. While substantial progress has been made in developing spectral libraries of known reflectance properties for specific plant physiological properties at the leaf and whole plant level, limitations in the application of remote imagery to production settings arise through the need for adequate ground-truthing of imagery to field conditions in the absence of atmospheric distortion. Field spectroradiometric measurements of individual cotton (Gossypium hirsutum, L.) leaves, dry soil and wet soil were used to create hyperspectral images with known radiometric, geometric and spatial properties. Fourier series were used to interpolate the spectrum of individual leaves and soil for subsequent generation of synthetic mixed spectra. The mixtures of hyperspectral profiles were built from the randomized linear summation of soil and vegetation spectra. The resulting spectra were assigned to pixels that simulate a resolution of 1m x 1m size. The dimensions of the generated plots were 18 m x 20 m, similar to the size of the field plots from which the spectral measurements were taken. From these synthetic images, we will explore the impact of leaf angle, soil reflectance and atmospheric distortion on canopy reflectance from aerial imagery.