Enhanced Fiducial Marker Based Precise Landing for Quadrotors

Abstract

Fiducial markers are landmarks typically used for pose estimation. However, the detection rate of existing fiducial marker based approaches is relatively low. To speed up the marker detection without sacrificing accuracy or robustness, this paper proposes a novel enhanced fiducial marker detection-tracking algorithm wherein the ROI (region of interest) based kernelized correlation filter (KCF) tracking strategy and the existing Apriltag2 algorithm are integrated together to form a new detection-tracking framework. More specifically, by selecting the appropriate ROI aera from the fiducial marker, the KCF tracking algorithm is employed to track the ROI area and the AprilTag2 algorithm is adopted to detect the selected area to obtain the pose estimation. Accordingly, the relative pose and the IMU measurements are fused in an Extended Kalman Filter (EKF) to obtain the full state estimation. On this basis, a nonlinear cascade P-PID controller is introduced to achieve the precise landing task. Comparative experiments are conducted to show that the detection rate of the proposed approach is up to 3-7 times faster than other methods without any parallel computation technique. In addition, the accuracy of the real-time landing is less than 5cm.