An N-dimensional approach towards object based classification of remotely sensed imagery

TL;DR

This paper introduces an N-dimensional object-based classification method using hierarchical support vector machines, combining spatial and spectral data with automatic optimization techniques to improve land cover mapping accuracy.

Contribution

It presents a novel N-dimensional hierarchical SVM approach incorporating cellular automata and genetic algorithms for enhanced remote sensing image classification.

Findings

Higher classification accuracy achieved compared to pixel-based methods

Effective reduction of class overlap through combined spatial and spectral data

Automatic kernel and parameter tuning improves model performance

Abstract

Remote sensing techniques are widely used for land cover classification and urban analysis. The availability of high resolution remote sensing imagery limits the level of classification accuracy attainable from pixel-based approach. In this paper object-based classification scheme based on a hierarchical support vector machine is introduced. By combining spatial and spectral information, the amount of overlap between classes can be decreased; thereby yielding higher classification accuracy and more accurate land cover maps. We have adopted certain automatic approaches based on the advanced techniques as Cellular automata and Genetic Algorithm for kernel and tuning parameter selection. Performance evaluation of the proposed methodology in comparison with the existing approaches is performed with reference to the Bhopal city study area.