Support vector machine for classification of various crop using high resolution LISS-IV imagery

Pradeep Kumar; Rajendra Prasad; Dileep Kumar Gupta; Varun Narayan Mishra; Arti Choudhary

Pradeep Kumar Department of Physics, Indian Institute of Technology (BHU), Varanasi, India
Rajendra Prasad Department of Physics, Indian Institute of Technology (BHU), Varanasi, India
Dileep Kumar Gupta Department of Physics, Indian Institute of Technology (BHU), Varanasi, India
Varun Narayan Mishra Department of Physics, Indian Institute of Technology (BHU), Varanasi, India
Arti Choudhary Department of Civil Engineering, Indian Institute of Technology, Guwahati, India

Abstract

The Resourcesat-2 is an exceedingly suitable satellite with its improved features and capabilities for crop classification studies. Data from one of its sensors, the Linear Imaging Self-Scanning (LISS-IV), which has a spatial resolution of 5.8 m, was used for the classification of various crop and non-crop in Varanasi district, Uttar Pradesh, India. The imagery was classified into classes of crop such as corn, linseed, lentil, mustard, barley, wheat, pigeon pea, sugarcane, pea and other crops and non-crop such as fallow land, sparse vegetation, dense vegetation, sand, built up, andwater classes.
The overall accuracies achieved by support vector machine (SVM) with polynomial of degrees 3, 4, 5 and 6 were 87.77%, 87.96%, 88.15% and 88.15% and kappa (Â·) 0.8686, 0.8706, 0.8726 and 0.8726 respectively. Results derived from SVM with different degree polynomialswere validated with the ground truth information acquired by the field visit on 6 April 2013.