Noise-Aware Event-Triggered Geographic Localization for UAV CMOS Vision Sensors
Year: 2026
Type: article
Abstract: This letter presents geographic region of interest localization (GeoROI-Loc), a software-assisted uncrewed aerial vehicle (UAV) localization framework that combines lightweight interframe tracking with event-triggered geographic correction. A physics-inspired complementary metal–oxide–semiconductor degradation protocol—including photon shot noise, readout noise, photo-response nonuniformity, and dark-signal nonuniformity—is introduced to investigate how sensor-output quality affects localization performance. GeoROI-Loc invokes CAMP-RoMa-PnP-based absolute correction only when accumulated motion or tracking uncertainty exceeds predefined thresholds. Under clean sensor conditions, GeoROI-Loc achieves an RMSE of 11.43 m, an Acc@1m of 42.35%, and an Acc@5m of 88.85%, with a latency of 1.23 s. Compared with the strongest competing method, it reduces RMSE by 47.08% and latency by 64.76%. These results demonstrate that GeoROI-Loc substantially improves localization accuracy and computational efficiency, highlighting its potential for resource-constrained UAV localization.
Source: IEEE Sensors Letters
Institution: Nanjing University of Aeronautics and Astronautics
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $2,645
OpenAlex ID: W7203664513
