Classification of radio elements using mutual information: A tool for geological mapping

A broad range of current airborne gamma ray spectrometry (AGRS) applications involve environmental mapping and mineral exploration. One common goal for such applications is the development of an algorithm for reliable on line classification of radio elements. In this paper, we propose the concept of maximization of correlated information as the similarity measure for classification. In order to achieve this similarity measure, we have developed an algorithm using the concept of minimization of mutual information, which is computationally faster, and requiring less memory than the hierarchical agglomerative clustering (HAC) method. The minimization of mutual information is achieved by maximizing the correlated information of the correlation matrix. The correlated information is maximized by the determination of its lower bound using the technique of determinant inequalities developed by us. We demonstrate the robustness of our results using mutual information and its superiority over that of Ward's method of minimum variance for the aerial survey carried out in central India.