Paper Title
Semi Supervised SVM Learning Technique for Land-Cover Mapping Using Spectral Information

Abstract - In this article, an approach using semisupervised support vector machines (S3VMs) is investigated for the problem of multispectral image classification of remote sensing images. S3VMs are developed using the concept of maximizing margin on both labeled and unlabeled samples. The effectiveness of the proposed technique is first exhibited on two labeled remote sensing (RS) data represented in terms of feature vectors and then mapping different land cover types in RS imagery. Investigation on these datasets shows that employing additional unlabeled points alongwith original ground truth samples increases the accuracy level. Comparison is made with the existing methods in terms of number of training examples, kappa value, accuracy and quantitative cluster validity indices. Keywords - Landcover mapping, Remote sensing satellite images, Semisupervised SVM, RBF Kernel function, Quantitative cluster quality indices.