Research on an Improved Stepwise Feature Matching Algorithm for UAV Indoor Localization

Indoor unmanned aerial vehicles (UAVs) localization faces challenges such as insufficient feature matches, high false match rates, and low accuracy under low-light conditions, viewpoint changes, and occlusions. To address these issues, we propose an adaptive FAST (Features from Accelerated Segment Test) corner detection thresholding method that adapts to indoor lighting and texture variations. Building upon the FAPP algorithm, our approach optimizes threshold selection during fine matching when correct and incorrect matching pairs are unevenly distributed. The proposed algorithm demonstrates superior robustness and efficiency in feature matching. In actual indoor UAV flight experiments, the feature matching rate reached 96.5%, and the indoor localization accuracy reached 0.02 meters. Compared to existing literature, both the average correct matching rate and localization accuracy were significantly improved, effectively solving the problem of precise indoor UAV localization.