Line Feature Correspondence between Object Space and Image Space

Abstract

In recent years, linear features have been gradually integrated into photogrammetry-related applications to estimate orientation parameters and facilitate object reconstruction, to name just a few. Yet, the correspondence of the conjugate linear features between different spaces or coordinate systems remains crucial to the goal of photogrammetric automation. When establishing exterior orientation of images on the line feature basis, the identifications and measurements of control line features must be involved. In this study, the authors develop an approach for effectively matching line features between object space and image space. The proposed algorithms start from projecting 3D line features into 2D image space employing collinearity equations with approximate orientation parameters. Then, the candidate lines detected from the images are chosen and matched with the 3D line features by imposing, in a sequential manner, geometric constraints that include angle, distance, reference point, and imaging geometry checks. Furthermore, a two-stage matching strategy, where the outcome of the first matching stage by partial 3D line features provides confined matching candidates for the second matching stage, is found effective in lowering the computational load, especially when faced with a great amount of 3D line features. Preliminary tests demonstrate successful, as well as satisfactory, 3D to 2D line feature correspondences using the proposed approach and strategy.