Understanding Temporal Semantic Stability in Open-Vocabulary UAV Perception through Metric 3D Fusion
Understanding Temporal Semantic Stability in Open-Vocabulary UAV Perception through Metric 3D Fusion
This research investigates temporal semantic stability in UAV perception by associating frame-wise predictions with persistent world-space locations through metric 3D fusion. The study introduces a voxel-level evaluation framework that characterizes semantic agreement, belief drift, and observation persistence, revealing that high aggregate world-space agreement can overstate temporal stability when many locations have limited repeated-observation support.
This paper addresses the challenge of temporal semantic stability in UAV perception by introducing a voxel-level evaluation framework that assesses semantic agreement, belief drift, and observation persistence. The findings highlight the importance of considering observation support when evaluating long-horizon semantic reliability.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
Source：arXiv Robotics — research abstracts · arxiv.org