Point cloud quality metrics for incremental image-based 3D reconstruction

• DOI: 10.1007/s11042-025-20596-6
• Publication Date: 2025-01-08T06:21:42Z
• Authors: Kai A. Neumann; Reimar Tausch; Hasan Kutlu; Arjan Kuijper; Pedro Santos; Dieter Fellner
• Keywords: Research Line: Computer vision (CV); LTA: Generation, capture, processing, and output of images and 3D models; Quality metrics; Point clouds; 3D Reconstruction; Branche: Cultural and Creative Economy; Next-best view planning

Abstract Image-based 3D reconstruction is a powerful method for accurately reconstructing an object’s geometry and texture from images. A crucial factor the accuracy completeness of resulting reconstructed model choice poses capturing images, which called view planning. One possible planning strategy uses iterative feedback loop that switches between incremental to autonomously digitize object without prior knowledge. However, this approach requires identifying parts are “poorly reconstructed” thus would benefit being part additional This work explores use point cloud quality metrics provide by comprehensively comparing set existing newly introduced in terms their time-dependent behavior, similarity, applicability Among proposed introduces Reconstruction Quality Feedback (RQF) , shows significantly improved performance simulations when used The effectiveness RQF also demonstrated real objects on autonomous robotic digitization system.