Detection of holes in 3D architectural models using shape classification based Bubblegum algorithm

Global digitalization with connectivity and smart devices have added an extra dimension to virtual experiences of heritage sites. However usage of crowd-sourced images may not be appropriate for optimized 3D reconstruction leading to holes. We propose a hole detection algorithm that detects holes at the point cloud phase of the 3D reconstruction pipeline. Most of the research work reported for hole detection uses meshes of the 3D models. In our algorithm, we detect holes and shapes using point clouds giving optimization in terms of computational time. Further on, the shape classification leads to specific structure based geometry which can be used to suggest an appropriate hole filling methodology. We tested our results on point clouds of Banashankari and Kalmeshwar temples on a system with 128 GB RAM, Intel Xeon running Ubuntu 16.04. (C) 2020 The Authors. Published by Elsevier B.V.