Globally Optimal Camera Orientation Estimation from Line Correspondences by BnB algorithm

This letter is concerned with the problem of estimating camera orientation from a set 2D/3D line correspondences, which major part Perspective-n-Line (PnL) problem. There are some cases that usually occur in real applications for PnL: input correspondences corrupted by mismatches (a.k.a. outlier correspondences). The RANdom SAmple Consensus (RANSAC) algorithm de facto standard solving outlier-contaminated PnL problems. However, RANSAC non-deterministic algorithm, means it produces reasonable result only certain probability. Therefore, could obtain certifiably optimal solution data matter priority safety-critical applications. In this letter, we take big step towards goal investigating globally estimation algorithms. Firstly, decouple rotation and translation considering geometrical property Branch-and-Bound (BnB) applied searches entire space to orientation. To investigate performance our method, tested proposed on both synthetic data, results show algorithms can more robust than several state-of-the-art methods.