2026 Volume 7 Issue 1 Pages 306-314
The quality of point cloud data generated by Structure from Motion and Multi-View Stereo (SfM/MVS) is considered to be influenced by illumination conditions; however, quantitative investigations of this effect remain limited. This study aims to clarify the minimum illumination required for successful SfM/MVS reconstruction and the illumination conditions necessary to obtain high-quality point cloud data. Video data were acquired in a classroom environment under controlled illumination levels and used as input for SfM/MVS analysis. In this study, successful reconstruction was defined as the condition in which the entire trajectory of the captured video, from start to end, was utilized in the analysis. The results indicate that, within the scope of this study, SfM/MVS reconstruction is achievable under illumination levels of 346 lx or higher when bidirectional image acquisition is performed. In addition, a decreasing trend in the number of feature points was observed with decreasing illumination. Furthermore, under illumination gradient conditions, it was suggested that point cloud quality improves when image acquisition is performed from lower to higher illumination regions.
