This paper presents a new algorithm for the automatic registration of airborne LiDAR data with aerial images using building corner features as registration primitives. First

three-dimensional building outlines are directly extracted from LiDAR points and building corner features which consist of two orthogonal straight lines are obtained by the regularization of three-dimensional building outlines. Straight lines are also extracted from every aerial image. Second

the building corner features are projected onto aerial images and corresponding image corner features are determined using the similarity measures. Lastly

the exterior orientation parameters are refined by bundle adjustment using the corner points of corner features as control points. Iteration strategy is adopted to obtain optimal results. The main advantage of the proposed algorithm is that the three-dimensional building outlines are extracted directly from LiDAR points without transforming LiDAR points into range image or intensity image

and therefore there are no interpolation errors. The experimental results show that the proposed algorithm can obtain more accurate results in comparison with the registration method based on LiDAR intensity image.