Research on 3D reconstruction of small size objects using structure from motion photogrammetry via smartphone images

Abstract

According to the recent developments on image acquisition via relatively low-cost devices such as smartphones or tablets, researches on Structure-from-Motion (SfM) photogrammetry-based 3 dimensional (3D) model reconstruction become more popular and wide frequent practice especially for study areas in geology/geomorphology, historical heritage, forestry, etc. Thus, this paper demonstrates accuracy assessment of SfM photogrammetry-based 3D model reconstruction of small size objects by altering number of the source imagery captured by smartphone and represents change detection of the generated datasets as (i) cloud-to-cloud and (ii) mesh-to-cloud comparisons. The number of images was decreased as 25% of each datasets belonging three different detail types of small size objects as the Safranbolu miniature house (SMH), a trinket made of a combination of historical buildings in Rome (HBR), and a wooden object (WO). A total of 12 datasets were generated and 9 cloud-to-cloud, and 21 mesh-to-cloud comparisons were performed. Since the obtained results show that the quality of 3D models of objects varies according to their shapes and sizes, change detection analyses show that the detail level of the objects are highly correlated with the resultant model accuracy.