Mixed reality head mounted displays for enhanced indoor point cloud segmentation with virtual seeds

Mixed Reality(MR)Head Mounted Displays(HMDs)offer a hitherto underutilized set of advantages compared to conventional 3D scanners.These benefits,inherent to MR-HMDs albeit not originally intended for such appli-cations,encompass the freedom of hand movement,hand tracking capabilities,and real-time mesh visualization.This study leverages these attributes to enhance indoor scanning process.The primary innovation lies in the con-ceptualization of manual-positioned MR virtual seeds for the purpose of indoor point cloud segmentation via a region-growing approach.The proposed methodology is effectively implemented using the HoloLens 2 platform.An application is designed to enable the remote placement of virtual tags based on the user’s visual focus on the MR-HMD display.This non-intrusive interface is further enriched with expedited tag saving and deletion functionalities,as well as augmented tag visualization through overlaying them on real-world objects.To assess the practicality of the proposed method,a comprehensive real-world case study spanning an area of 330 s^(2) is conducted.Remarkably,the survey demonstrates remarkable efficiency,with 20 virtual tags swiftly deployed,each requiring a mere 2 s for precise positioning.Subsequently,these virtual tags are employed as seeds in a region-growing algorithm for point cloud segmentation.The accuracy of virtual tag positioning is found to be exceptional,with an average error of 2.4±1.8 cm.Importantly,the user experience is significantly enhanced,leading to improved seed positioning and,consequently,more accurate final segmentation results.

Comparison of heuristic and deep learning-based methods for ground classification from aerial point clouds

The automatic definition of the ground from 3D point clouds has been a common process for the last two decades,with many different approaches and applications that can be found in a vast literature.This paper presents a comparison of three different methodological concepts for ground classification,in order to establish the advantages and drawbacks of each method.First,a heuristic method,based on previous knowledge of the geometry and context of the 3D data.Secondly,a Deep Convolutional Network based on SegNet that classifies 2D images generated from the 3D point cloud.Finally,the third method applies a Deep Learning classification based on PointNet,which takes 3D points directly as inputs.To validate each method and compare them,public and labelled point clouds from the Actueel Hoogtebestand Nederland dataset are employed.Furthermore,the three methods are validated against the ISPRS 3D Semantic Labeling Contest benchmark.The results obtained show that the deep learning-based approaches outperform the heuristic method,with F-scores above 96%.The best results were obtained using a shallower version of SegNet,with F-score above 97%.