Search papers, labs, and topics across Lattice.
This study benchmarks five monocular SLAM systems using aerial nadir video, which presents unique challenges due to weak geometric constraints and perceptual aliasing. The evaluation reveals significant variability in performance based on environmental conditions and trajectory scale, with MASt3R-SLAM achieving the best results on local DJI flights and DROID-SLAM performing best overall across various runs. However, the findings indicate that existing monocular SLAM methods struggle with vertical positioning and maintaining global trajectory shape, highlighting their limitations for reliable visual-only aerial navigation.
Monocular SLAM systems falter in high-altitude aerial navigation, with no single method consistently preserving trajectory shape or achieving reliable vertical positioning.
Aerial nadir video combines weak geometric constraints with severe perceptual aliasing, making it a difficult regime for monocular SLAM. We benchmark five monocular SLAM systems on local UAV flights, synthetic city-scale imagery, and long-range aerial sequences. To isolate visual performance, we provide no inertial or GNSS aiding. Performance varies strongly with environment and trajectory scale: MASt3R-SLAM achieves the lowest mean horizontal MAE on the five DJI flights (0.53\% of reference path length), whereas no system consistently preserves global trajectory shape on the long GES and ALTO sequences. Overall, DROID-SLAM performs best, averaging 2.88\% of reference path length across completed runs. Vertical position remains poor, and large-area trajectories remain highly distorted despite loop-closure capability. Current monocular SLAM methods are by themselves therefore insufficient for reliable visual-only aerial navigation.
