A monocular visual single-axis rotation measurement method of multi-rudders based on cooperative targets

Abstract

To tackle the problem that the rotation angle of multiple rudders needs to be measured in real time and high precision, a measurement method based on cooperative targets is proposed in this paper. Firstly, to detect and encode multiple similar rudders with sparse feature points in complex backgrounds, we design a novel visual cooperative target with encoding information, and study a cooperative target localization algorithm based on contour extraction as well as a corner point extraction algorithm based on a checkerboard grid. Then, through the PnP algorithms and numerous corner points extracted from the cooperative target, an unconstrained optimization model is developed based on the re-projection error, and the Levenberg-Marquard (LM) algorithm is applied to estimate the rotation angle. Finally, an experiment platform is set up to evaluate the accuracy and speed of the algorithm. Compared with the existing methods, the experimental results show that the proposed cooperative target exhibits favorable localization under various typical experimental environments with a detection success rate better than 98%, meanwhile, the actual measurement accuracy of the angle estimation model is better than 0.05° and the average execution time of each measurement is less than 20 ms. Moreover, the proposed method is easy to install and operate in actual measurement, and can achieve simultaneous measurement of the spatial rotation angle of multiple rudders, which is well suited to actual industrial production needs.