Abstract
Traditional classifying and searching systems have required special features of data from manual input. To overcome this undesirable task, a method which automatically creates and utilizes a "Critical Point Graph (CPG)" as an index of volume data has been proposed. In order to achieve suitable search and classification results, it is required to automatically calculate similarity between CPGs from a set of selected volume data. We introduce a new feature-matching algorithm to make an exact correspondence of positions between two CPGs. This is an extension of "Triangular Matching Algorithm" which is usually employed in the field of 2D fingerprint matching. In order to evaluate the effectiveness, we evaluated the influence of data resolution and arrangement in our proposed CPG based method. We confirmed the effectiveness and usefulness of our proposed CPG based method when applied to numerically constructed weather volume data sets. From computational experiments, our new CPG method has shown suitable ability for similarity calculation between two dissimilar volume data.
