ANFIS based Information Extraction using K-means Clustering for Application in Satellite Images

Information extraction from satellite images is a challenging task. This is because of the associated uncertainty arising out of improper capture and subsequent transfer. Fuzzy systems are suitable for such applications because of the fact that these have the ability to capture minute variations in the patterns presented. Fuzzy systems are expert decision making tools that require support from Artificial Neural Network (ANN) for inference generation. This leads to the formation of NeuroFuzzy system (NFS). A NFS requires certain apriori information for making appropriate decision. Apriori knowledge can be provided manually but it becomes tedious, hence certain computational approaches are required. In this paper, the focus is given on the development of an information extraction system based on K-means clustering (KMC) and ANN and an adaptive neurofuzzy inference system (ANFIS) based system with the same purpose to achieve enhanced performance as compared to each other. We specially deal with an ANFIS aided by KMC for use with information extraction from satellite images. Experimental results show that such system is fully automatic and effective in dealing with information extraction from river images with forest and sand distribution along its banks. General Terms Adaptive Neuro Fuzzy System, Neuro Fuzzy System, Artificial Neural Network, K-means clustering.