Probabilistic land cover classification approach toward knowledge-based satellite data interpretations

The recognition of concepts that we human beings are able to locate within satellite imagery requires analysis based on the particular context using knowledge. In this paper, we present a supervised pixel-based classification approach toward utilization of the classification results in knowledgebased satellite data interpretation system. The proposed approach is based upon a generative model, which is able to output the classification results with their probabilities and subsequently utilize them in detailed analysis. The experiment of classification was performed to demonstrate characteristics of the approach.