UAV hovering pose detection via binocular vision system based on improved SAC-IA+ICP point cloud registration

Abstract

In this paper, to solve the problem of quantitative evaluation of the stability of the unmanned aerial vehicle (UAV) while hovering, a binocular vision relative pose measurement system that can be used for remote control is developed and a characteristic target adapted to the system is designed. Moreover, a feature point extraction algorithm based on known feature constraints and a fast feature point matching algorithm based on projection shape invariant constraints are proposed for fast and accurate extraction and reconstruction of feature point sets. The point cloud registration method of SAC-IA + ICP is used to solve the pose and orientation of the characteristic target. It is verified by simulation that the orientation and angle of the target have registration accuracies of 1.1078mm and 0.0463°, respectively. The binocular system is experimentally analyzed using the laser tracker and T-Mac measurement as the true value. Compared with the optical flow method, the former time is shortened by 60%, the Euclidean displacement measurement error is reduced by 83%, and the attitude angle deflection measurement error is reduced by 87%. Then, the hovering state of the UAV is measured using the binocular system; and the obtained experimental measurement results of the system are true and reliable.