Helipad Pose Estimation Using Intelligent Helipad and LiDAR Camera Fusion (IHLCF)

摘要

Pose estimation has evolved into a beneficial concept in autonomous systems. It refers to the techniques used by computers to detect and quantify certain features in an image. The present work proposes modified helipad intelligent detection and pose estimation, using a fusion of camera and LiDAR. The image data are first collected using Otsu thresholding through the downward drone camera and converted to a binary image. Next, Boundary Parametric Ellipse Fitting (BPEF) algorithm is employed to detect circles, which will turn into ellipses when there is a tangential distortion in an image. Then, Ellipses Region of Interest (EROI) is extracted from the images via the potential circles. The algorithm uses a modified version of the helipad with an arrow sign located outside of the helipad's circle. The arrow's centroid point is located on the axial line, which horizontally splits the word "H" and passes the word's centroid. Hence, using the proffering over-the-line-and-between-ellipses-check technique, potential arrows are extracted. A Support Vector Machine (SVM) is then trained to detect the helipad over 400 images of the word "H" and Arrow patterns. The "H" and the Arrow corners are detected and localized in the following phase. The projected LiDAR data is followingly utilized to find the corners depth information. Finally, the translational and rotational pose components are projected to obtain the corners' coordinates and the rigid body transformation. Software-in-the-Loop (SIL) is used to assess the method accurately. The experimental setup is tuned so that the drone stays motionless over the landing platform and conducts the pose estimation. The method was compared with the AprilTag Detection Algorithm (ATDA). A statistical Root Mean Square Error (RMS) is also used to gauge the accuracy of the proffered method. The analysis results confirmed a notable improvement in rotational and translational estimations.