A Technique for Classification of High Resolution Satellite Images Using Object-based Segmentation

Segmentation and classification of satellite imagery are a challenging problem due to the fact that it is not feasible to execute this task on a pixel-by-pixel basis.This paper proposes an efficient image classification technique for satellite images with the aid of KFCM and artificial neural network. The proposed classification technique is made of four phases namely, pre-processing, fragmentation and final classification using neural network. In pre-processing stage, the input image is put to a series of preprocessing phases which includes mean filtering, contrast adjustment and laplacian filter. The preprocessing has the effect of modifying the input image into suitable image for fragmentation. After the preprocessing, the image is fragmented by means of the kernel based fuzzy c-means(KFCM) clustering algorithm. The resultant image obtained is segmented into clusters. NN is trained according to the data given. The resultant image is given as input to the trained NN, which classifies the satellite images into road, building and vegetation regions according to the trained data and pixel values.The experimental results have demonstrated the effectiveness of the proposed classification technique in classifying into road, building and vegetation regions. The experimentation is carried out using the satellite images and the analysis ensures that the performance of the proposed technique is improved compared with k-means clustering algorithm.