Comparative analysis of classification techniques for crop classification using airborne hyperspectral data

Crop classification using high-dimensional and high-resolution data is a challenging task. Though a large number of classes can be obtained from the hyperspectral data, the "curse of dimensionality" causes the classification accuracy to be less than the expected value. A minimum noise transform has been applied to the data in this work, to reduce dimensionality and improve classification accuracy. This paper compares the different methods of supervised and unsupervised classification for the identification of different crops in a field. The results showed that it is better to use supervised methods over unsupervised as they yield better classification accuracy and kappa coefficient.