Абстрактный

Spatial Information Based Image Classification Using Support Vector Machine

P.Jeevitha, Dr. P. Ganesh Kumar

Segmentation and classification of images are major task of remote sensing. Active learning approach in combination with machine learning is recently developing for classification of such satellite images. It optimizes the task of collecting the training set samples used for classification. In this paper, a support vector machine is developed for image classification without using active learning approach and with using active learning approach and their performance is compared. Present classification approaches concentrates on spectral features of the image. But for SVM with active learning, we adopt two criteria: spectral and spatial. Experiments are made using high resolution spectral images and the effectiveness of active learning algorithm using both spectral and spatial criteria is observed.