On benchmarking camera calibration and multi-view stereo for high resolution imagery

In this paper we want to start the discussion on whether image based 3-D modelling techniques can possibly be used to replace LIDAR systems for outdoor 3D data acquisition. Two main issues have to be addressed in this context: (i) camera calibration (internal and external) and (ii) dense multi-view stereo. To investigate both, we have acquired test data from outdoor scenes both with LIDAR and cameras. Using the LIDAR data as reference we estimated the ground-truth for several scenes. Evaluation sets are prepared to evaluate different aspects of 3D model building. These are: (i) pose estimation and multi-view stereo with known internal camera parameters; (ii) camera calibration and multi-view stereo with the raw images as the only input and (iii) multi-view stereo.