Semantic segmentation of point clouds of ancient buildings based on weak supervision

Abstract

Abstract Semantic segmentation of point cloud of ancient buildings plays an important role in heritage building information modeling (HBIM). Since the point cloud annotation task of ancient architecture point cloud is characterized by strong specialization and large workload, which greatly restricts the application of point cloud semantic segmentation technology in the field of ancient architecture, this paper researches on the semantic segmentation method based on weak supervision for ancient architecture point cloud. Aiming at the problem of small differences between classes of ancient architectural components, this paper introduces a self-attention mechanism, which can effectively distinguish similar components. We also explore the insufficiency of position encoding in baseline to construct a high-precision point cloud semantic segmentation network model for ancient buildings. We call it SQN-DLA. using only 0.1% of the annotations in our homemade dataset and the public dataset ArCH, the mIoU reaches 66.02% and 58.03%, respectively, which is improved by 3.51% and 3.91% compared with baseline, respectively.