Application of EO-1 Hyperion Data for Mapping and Discrimination of Agricultural Crops

Remote sensing offers an efficient and reliable means of collecting the information required for mapping, assessing and monitoring of agricultural crop conditions and production. Recent advances in remote sensing technology have led to the development of hyperspectral remote sensing imaging devices which can obtain high-resolution radiance data. This study evaluates the potential of the hyperspectral data in discrimination and mapping of agricultural crops using EO-1 Hyperion hyperspectral image over the Thalasseri Taluk, Kerala, India. Five agricultural crops such as arecanut, banana, cashew, coconut and rubber were considered for the study. The EO-1 was pre-processed using minimum noise fraction (MNF) transform to reduce the atmospheric effects on the imagery. Support vector machine classification and minimum distance classification were applied in order to perform image data classification based on different crops. The optimum wavelengths suitable for crop discrimination were derived by analysing the spectral reflectance curve as well as by using the techniques such as stepwise discriminant analysis and partial least square regression (PLSR). This study establishes that the Hyperion bands 53, 56, 62, 74, 79 and 84 are suitable for crop-type discrimination. The support vector machine classification is suitable for mapping the crops from Hyperion imagery with a higher accuracy of about 80% and above.