A spatial and spectral feature based approach for classification of crops using techniques based on GLCM and SVM

Abstract

This paper highlights the study regarding the classification of crop types using the techniques based on Gray Level Co-occurrence Matrix (GLCM) and support vector machine (SVM). The dataset used was from IRS-LISS IV sensor with 5.8 m spatial resolution having three spectral bands of date 4-October 2014 for our chosen location at 20°07′13.5″N 75°23′05.3″E. Classification of all three bands followed by classification of GLCM measures (of all three bands) was accomplished by using Support Vector Machine classifier with Radial Basis Function. The accuracy of classification obtained from GLCM was 90.29% with the Kappa coefficient 0.88 whereas the corresponding values obtained from three band classification were 86.04% and 0.83, indicating the superiority of the GLCM-based approach.