Neighborhood-Assisted Point Cloud Inference

Résumé

Abstract This research introduces a novel method for efficient 3D reconstruction, focusing on accelerating point cloud inference using silhouettes. Addressing the limitations of traditional reconstruction techniques, this approach streamlines the generation of point clouds by leveraging geometric data from silhouettes and incorporating a neighborhood-assisted algorithm for rapid depth calculation. The method involves capturing video of rotating objects, extracting silhouettes, and inferring pixel depths to form point clouds. This technique, distinct from conventional stereoscopy and active vision approaches, emphasizes efficiency and simplicity. Experimental results demonstrate the method's effectiveness in significantly enhancing the speed of point cloud generation without compromising accuracy. The experiments, conducted with a variety of objects, validate the method's precision and temporal efficiency. This study has broad implications, offering a practical solution for rapid 3D modeling in fields such as virtual reality, cultural heritage, and medical imaging. It marks a significant advancement in 3D reconstruction, providing a foundation for further research aimed at optimizing computational efficiency and integrating with more detailed reconstruction methods.

Citer ce document

Accès au document

Texte intégral en lecture en ligne, réservé aux abonnés SPHAERO et aux membres de l'institution. Se connecter

Voir l'article sur le site de la revue

Auteur(s)

Statistiques

Consultations : 1

Téléchargements : 0