Main content area

Comparison of BRDF models with a fuzzy inference system for correction of bidirectional effects

Bryant, R., Qi, J., Moran, M.S., Ni, W.
Remote sensing of environment 2003 v.88 no.3 pp. 221
image analysis, reflectance, prediction, model validation, fuzzy logic, remote sensing, cotton, corn, plant density, precision agriculture, Texas, California
With the advances in computing and imaging technology, the field of precision agriculture is rapidly becoming a practical means for farm management. An important step in the delivery of highly accurate images for farm managers is the within-image correction for viewing geometry effects. Reflected light on an imaging sensor is influenced by properties of view zenith angle, solar zenith angle, and relative azimuth. There are a number of models that describe this effect termed the bidirectional reflectance distribution function (BRDF) or more generically “viewing geometry effects.” In this paper, we compared three BRDF models (Roujean, Shibayama-Wiegand, and Dymond-Qi) with a fuzzy inference system (FIS) for three data sets for correction of geometric effects. One data set consisted of ground data collected at different viewing angles of a cotton crop. Another data set included six aircraft images of a corn plot in a different part of each image. The final data set was an aerial image of a planting density experiment of cotton. All the models performed reasonably well, but the FIS was the most consistent predictor of BRDF for all three data sets. For the ground data set, R2 statistics for predicting the reflectance based on the trained models ranged from 0.53 to 0.93 for the BRDF models and from 0.94 to 0.97 for the FIS.