Camera path design and evaluation in ground-based image acquisition for Structure from Motion and 3D modeling

The Structure from Motion (SfM) is a useful approach to generate 3D data for various, such as infrastructure inspections, disaster monitoring and archaeological surveys. Images taken from random viewpoints can be used in the SfM processing. However, a plenty of processing time is required to detect feature points and corresponded points from images in the Scale-Invariant Feature Transform (SIFT) procedure. When the number of images for the SfM is minimized, the efficiency in SfM processing can be improved. Moreover, the minimization of the number of images can achieve a shorter flight time in U AV operation. Therefore, we focus on a camera path design is required to improve the efficiency in images acquisition for SfM processing. In this research, we proposed a camera path design to improve the efficiency in the SfM for a reconstruction of complex objects. We conducted experiments to verify our camera path designs for several image acquisition patterns. Through our experiments, we evaluated our camera path designs with the number of images for the SfM and point cloud density.