An autonomous alignment method for UAVs with wind turbine blades based on 3D vision guidance
TL;DR
A quadrotor UAV positive alignment algorithm design for wind turbine blades based on RGB-D perception, contour skeletonization and robust plane fitting can provide the fan blades independent inspection and defect detection stability of geometric sense.
Abstract
Aiming at the problem that the camera line of sight is difficult to stabilize the blade surface in the autonomous inspection of wind turbine blades by unmanned aerial Vehicle (UAV), this paper proposes a quadrotor UAV positive alignment algorithm design for wind turbine blades based on RGB-D perception, contour skeletonization and robust plane fitting. Firstly, the color image and depth map obtained by the RGB-D camera mounted on the UAV are preprocessed, and the candidate leaf regions are extracted by graying, filtering, Canny edge detection, contour screening and mask filling. Then Zhang-Suen thinning algorithm is used to obtain the central skeleton of the leaf, and the skeleton points are connected sorted and equally spaced sampled. The sampling points are back-projected into a 3D point cloud combined with the participating depth values in the camera, and the local plane of the leaf is estimated by RANSAC. The range, lateral, vertical, yaw and pitch errors are calculated according to the center of the fitting plane and the normal vector, and the UAV velocity commands are generated by the state machine and the band limited PD controller. In this paper, a simulation platform is constructed based on ROS, PX4 and Gazebo, and the blade mirror detection and positive pair perception link verification are completed. Simulation experimental results show that the system is successful for her image, this method can provide the fan blades independent inspection and defect detection stability of geometric sense.