A fuzzy-based threshold method applied in SIFT for visual navigation of small UAVs

In this paper, a fuzzy-based threshold method is proposed in image processing of SIFT which is used in visual navigation of small UAVs. By adjusting the thresholds of SIFT online fuzzily, it provides a stable number of features detected from aerial image sequence. The visual navigation estimates the motion of the small UAV by tracking and matching these features extracted from the consecutive image frames. Real experiments verify the stability and advantage of this method in comparison with the fixed and PID-based threshold methods. This proposed method is able to stabilize the number of features regardless of the changes of the aerial images.