Identifying and Mapping Stripe Rust in Winter Wheat using Multi-temporal Airborne Hyperspectral Images

Plant disease incidence usually has a progress from light infection to severe prevalence. How to dynamically monitor this progress on a tempo-spatial scale has become a pressing issue for farmers and agricultural decision-making. In comparison with traditional field-point based detection and diagnosis, remote sensing techniques have provided cost effective tools for acquiring disease severities and corresponding spatial distribution. The primary objective of this study was to quickly identify and map stripe rust infections in winter wheat using multi-temporal airborne hyperspectral images from a Pushbroom Hyperspectral Imager (PIH) sensor developed by Chinese Academy of Sciences. Three PHI images were acquired from April to May in 2002 during the growing season of winter wheat. After comparatively analyzing the image and spectral properties between normal and diseased points in the PHI images, forty-five field sampling points were used to build a binary linear regression model, with a correlation coefficient (r) of 0.923 and the standard error of 0.108. Additional twenty points were utilized to validate the model and the coefficient of determination (R 2 ) reached 0.877, which showed that this model was encouraging. When applied the model to the three PHI images, winter wheat fields with different stripe rust infections were identified and mapped with five relative severity levels: normal, light, moderate, serious and very serious. The detection results indicated that stripe rust incidence was progressively severe from jointing to milky stage and it was more severe in the southern part than in the northern part, which were very coincident with the real field survey. © 2012 Friends Science Publishers