Search papers, labs, and topics across Lattice.
This paper introduces a Fast Reflectivity-Aware Next-Best-View (FRA-NBV) strategy that enhances 3D reconstruction in the presence of reflective surfaces, which typically hinder conventional methods. By identifying reflective regions through the spatial distribution of missing depth measurements and employing an online ellipsoid-based representation, the approach enables the selection of optimal sensor poses to mitigate depth loss. Experimental results show a significant improvement in reconstruction coverage across various industrial scenarios, highlighting the method's robustness and applicability in real-world settings.
Reflective surfaces no longer spell disaster for 3D reconstruction鈥擣RA-NBV significantly boosts coverage by intelligently adapting sensor poses in real-time.
Autonomous 3D reconstruction with depth sensors is strongly affected by reflective surfaces, which cause missing or unreliable measurements and reduce the effectiveness of conventional Next-Best-View (NBV) strategies. This limitation is particularly critical in industrial applications involving reflective components and low-cost, low-resolution depth sensing, where robustness to sensing failures is essential. This paper proposes a Fast Reflectivity-Aware Next-Best-View (FRA-NBV) strategy that explicitly addresses reflection-induced depth loss without relying on prior object models or assumptions on material reflectance, making it suitable for a wide range of industrial configurations. Reflective regions are identified from the spatial distribution of missing depth measurements and localized in three-dimensional space using an online ellipsoid-based representation of the object estimate. A recovery strategy then selects additional poses that modify the sensor's angle of incidence to improve the likelihood of reconstructing the affected regions. Experiments on objects with different geometric and reflective complexity demonstrate that the approach significantly improves reconstruction coverage under realistic industrial conditions.
