Classification of crops using multitemporal hyperion images

In this study, multi-temporal Hyperion and Landsat 8 images from Eastern Turkey are used to classify wheat, alfalfa and sainfoin (Onobrychis). 9 Hyperion and 9 Landsat 8 images acquired at similar dates from year 2013 are used for both same and cross date classification of crops. Ministry of Agriculture, Food and Live Stock´s farmer registration system (ÇKS, Çiftçi Kayıt Sistemi) records are used as ground truth after a refinement process. Preprocessing steps used for Hyperion data are presented. 100 samples from each class are input to the SVM and KNN classifiers. SVM classifier performs better than KNN. Landsat 8-launched in 2013- has better same date classification accuracy compared to Hyperion: 91.86% vs 88.94%. Cross date classification of crops with Hyperion and Landsat 8 images yielded lower cross date classification accuracy for both multispectral and hyperspectral images and only in certain periods such as harvest acceptable classification accuracies may be obtained.