Performance analysis of possisblistic fuzzy clustering and support vector machine in cotton crop classification

Cotton crop classification is found to be a significant task in crop management. Literature has exploited unsupervised fuzzy based classification and various vegetation indices for cotton crop classification. However, fuzzy based classification has negative effect on performance, because of inliers and outliers in the image. Hence, it is not reliable to investigate the performance of vegetation indices and for cotton crop classification. To overcome this drawback, this paper introduces possiblistic fuzzy c-means (PFCM) clustering for labeling the learning data and exploits support vector machine (SVM), which enables supervised learning, for cotton crop classification. Subsequently, five vegetation indices namely, simple ratio (SR), Normalized Difference Vegetation Index (NDVI), Soil Adjusted Vegetation Index (SAVI), Triangular Vegetation Index (TVI) and Transformed Normalized Difference Vegetation Index (TNDVI) are considered for investigation. LISS - III multi - spectral images of IRS - P6 sensors are acquired from Aurangabad region, India and they are subjected to experimental study. Three image sets are subjected to experimental investigation and the proposed classifier is compared with an existing classifier. The proposed classifier outperforms the existing classifier in all the image sets. Comparison in terms vegetation indices demonstrate that SR outperforms other vegetation indices by achieving 88.72%, 88.71% and 89.15% accuracy values for image sets 1, 2 and 3 respectively.