Automatic pose measurement of robotic drilling system based on zoom monocular vision

Abstract

Robotic drilling systems are attracting more attention due to their excellent processing accessibility and manufacturing flexibility, enabling in-situ processing of large-scale components. However, due to lack of feedback mechanism, it is difficult to achieve the required precision in hole-making positions. Although a laser tracker can be used to determine the positional deviation of the robotic drilling system with respect to the hole, accurately measuring the pose deviation still remains a formidable challenge. To tackle this issue, a sophisticated pose measurement system is proposed, which is composed of a zoom camera and a stereo cooperative target. To ensure that the pose can be effectively measured over a large range of distances, an automatic zoom calibration method based on Huber regression is proposed. Moreover, to establish the correspondence between the 3D target feature point coordinates and the 2D image feature coordinates, a novel automatic pose estimation algorithm is designed, which addresses the problem of matching failure for conventional pose estimation algorithms. Experimental results demonstrate that our pose measurement system can effectively complete the pose measurement task, with a measurement accuracy of 0.04° ranging from 3 to 7 m.