An Accurate Method for Solving the PnL Problem Based on the Reprojection Error

This paper proposes a new idea to solve the camera pose estimation problem using 2D/3D lines, which is known as the perspective-n-line (PnL) problem in computer vision. Although many successful solutions have been proposed, further improvement of the accuracy is still a challenge. Our main idea is to optimize the result twice based on the two forms of reprojection error to obtain the exact position after obtaining the initial value of the pose from polynomial equations. Simulation results show that the accuracy of our method is better than the state-of-the-art PnL algorithm in the presence of 2D line extraction errors and is computationally efficient. Real image test results show that our method can be used in a variety of situations.