Automated and Accurate Orientation of Large Unordered Image Datasets for Close-Range Cultural Heritage Data Recording

作者：Abdel Wahab Mohammed^*; Wenzel Konrad; Fritsch Dieter

来源：Photogrammetrie Fernerkundung Geoinformation, 2012, (6): 679-689.

DOI：10.1127/1432-8364/2012/0148

• automatic image orientation
• 3D reconstruction
• bundle adjustment
• close range
• cultural heritage

Reconstruction of image orientations and geometry from images is one of the basic tasks in photogrammetry and computer vision. A fully automated solution of this task in terrestrial applications is still pending in case of large unordered image datasets especially for close-range and/or low-cost applications. Current solutions require high computational efforts for image networks with high complexity and diversity regarding acquisition geometry. Unlike the methods suitable for landmark reconstruction from large-scale Internet image collections we focus on datasets where one cannot reduce the number of images without losing geometric information of the dataset. Within the paper, an automated pipeline for the reconstruction of reliable and precise camera orientation from unordered image datasets is presented. Results for a close-range cultural heritage application, the example of the Amsterdam project, are shown to demonstrate the performance of the presented pipeline for applications with low cost and high accuracy requirements.

• 出版日期2012