Image segmentation and classification are the major challenges to satellite imagery. Also, the identification of unique objects in the satellite image is a significant aspect in the application of remote sensing. Many satellite image classification techniques have been presented earlier. However, the accuracy of the image classification has to be further improved. So that, optimal artificial neural network with kernel-based fuzzy c-means ([Formula: see text]) clustering based satellite image classification is proposed in this paper. Initially, the images are segmented with the help of KFCM algorithm. Then, color features and gray level co-occurrence matrix (GLCM) features to be extracted from the segmented regions. Then, these extracted features are given to the OANN classifier. Based on these features, segmented regions are classified as building, road, shadow, and tree. To enhance the performance of the classifier, the weight values are optimally selected with the help of fruit fly algorithm. Simulation results show that the performance of proposed classifier outperforms that of the existing filters in terms of accuracy.
A hybrid interval type-2 semi-supervised possibilistic fuzzy c-means clustering and particle swarm optimization for satellite image analysis
