Accuracy Assessment of MOGA-SVM MethodComparing with Some Supervised and Unsupervised Classification Methods

In The problem of unsupervised classification of satellite image in number of homogeneous regions can be viewed as the task of clustering the pixels in the intensity space. This paper compares new method that combines recently proposed multi-objective fuzzy clustering scheme with support vector machine (SVM) classifier with four unsupervised and supervised methods like maximum likelihood (ML), SVM, fuzzy c-means (FCM), and k-means (KM). The multi-objective technique is first used to produce set of non-dominated solutions. The non-dominated set is then used to find some high-confidence points using fuzzy voting technique. The SVM classifier is thereafter trained by these high-confidence points. Finally, the remaining points are classified using the trained classifier. However, results demonstrating that supervised classification methods is better than unsupervised methods but new method (MOGA-SVM) shows the best result among other clustering methods. Moreover, TM satellite image of Qaem Shahr, Iran has been classified using the proposed technique to establish its utility.

Keywords:

Fuzzy clustering , multi-objective optimization (MOO) , support vector machine (SVM) , Genetic algorithm