KNN and SVM based Satellite Image Classification

Abstract: Remote sensing is the art of acquiring information about an object or area using machine or device that is not physically connected to the object. Geology, urban planning, soil assessment and land cover/land use are the different applications of remote sensing. Remote sensing is widely used for generation of classification map. Image classification is used to group the pixels present in an image into different classes. This paper presents Support Vector Machine (SVM) and K Nearest Neighbor (KNN) based classification system for Indian Remote Sensing (IRS) satellite images. The proposed system consists of image enhancement, segmentation, selection of training data and classification. For image enhancement adaptive histogram equalization is used. Image segmentation is carried out using K means and Fuzzy C Means (FCM) clustering. Linear Multi-Class SVM (MCSVM) and KNN techniques are used for classification of remotely sensed imagery.

Keywords: Classification, Clustering, K Nearest Neighbour, Multi-Class Support Vector Machine, Satellite imagery, Remote sensing.