Incremental Modelbase Updating: Learning new model sites

Each stage of the hierarchical clustering process involves one NP-complete clique finding process. In this paper, we present a method to add new mod* ,The cluster representatives at each stage are formed els automatically to an existing hierarchically organized .... by merging several child RPSDs below, and we modelbase. Given a new model, the model in the library closest to it is found by searching the tree, in a modification of the recognition strategy described in [1]. Next, we use the threshold values at each level of the hierarchy to learn the site of the new model in the organized tree. We present results of how the quality of the organized tree changes as the models are added to organized CAD libraries. know that merging is NP-hard. When a new model RPSD is added to the library, repeating the expensive, complete organization process for the library of size (N + 1) is prohibitive. Further, it discards the work already done in organizing the N model library. The addition, or the learning of the new model’s site in the library should be more efficient. The model adding process is described in Section 3.