Hybrid Photogrammetry Structure-from-Motion Systems for Scene Measurement and Analysis

Abstract

Many applications, such as robotics and modeling, require the accurate measurement of an object or scene. The field of metrology specializes in accurate measurement and there are many approaches to solve this problem. Photogrammetry is the most common technique. Targets (either coded or non-coded) are placed on an object, which is a time intensive operation. Photographs are taken and the target centers are triangulated to accurately compute the 3D positions 1. In computer vision, structure-from-motion (SfM) is used to obtain a 3D model from a collection of photographs. SfM involves a number of common computer vision methods 2. Photogrammetry and SfM have similar mathematics, but the goals diverge. Generally, photogrammetry is focused on obtaining very accurate 3D positions with low covariances. SfM favors visually dense reconstructions so that more knowledge of the scene can be obtained, but measurements are insufficient for metrology. There is work in SfM to obtain more accurate results3. Given the photogrammetry and SfM advances, an application harnessing the symbiotic relationship between these fields is now possible. The system uses photogrammetric information to enhance the accuracy of SfM. SfM provides a more complete reconstruction, where photogrammetry is used to correct camera parameters or scene points that have higher uncertainty. This procedure is less expensive, maintains the required metrological accuracy, and provides a dense reconstruction.

2.