अमूर्त

GABC based neuro-fuzzy classifier with multi kernel segmentation for satellite image classification

Ankayarkanni B, A Ezil Sam Leni

Segmentation combined with classification of satellite images is an important problem that could not be execute on the basis of pixel by pixel manner. Nowadays, several satellite image segmentation and classification methods are available. To overcome the difficulties in the existing techniques, a novel satellite image classification system using Multi Kernel Fuzzy C-Means Clustering (MKFCM) with optimal neuro-fuzzy classifier is proposed. The proposed work consists of three stages such as (i) satellite image segmentation (ii) feature vector generation and (iii) image classification. The input image is preprocessed, to make it suitable for segmentation. The image is segmented using Multi Kernel Fuzzy CMeans Clustering (MKFCM). Later the feature vector is created for further classification. The neurofuzzy classifier classifies the image into road, building and vegetation region. Here, the optimal rules are generated using hybridization of Genetic and Artificial Bee Colony (GABC) algorithm. Accuracy of the proposed method is better when compared with the existing method.