Improving Sparse 3D Models for Man-Made Environments Using Line-Based 3D Reconstruction

Abstract

Traditional Structure-from-Motion (SfM) approaches work well for richly textured scenes with a high number of distinctive feature points. Since man-made environments often contain textureless objects, the resulting point cloud suffers from a low density in corresponding scene parts. The missing 3D information heavily affects all kinds of subsequent post-processing tasks (e.g. meshing), and significantly decreases the visual appearance of the resulting 3D model. We propose a novel 3D reconstruction approach, which uses the output of conventional SfM pipelines to generate additional complementary 3D information, by exploiting line segments. We use appearance-less epipolar guided line matching to create a potentially large set of 3D line hypotheses, which are then verified using a global graph clustering procedure. We show that our proposed method outperforms the current state-of-the-art in terms of runtime and accuracy, as well as visual appearance of the resulting reconstructions.

Hierarchical line matching based on Line–Junction–Line structure descriptor and local homography estimation

“…Objects in real scenes, however, can be easily outlined by line segments, especially for man-made scenes. This indicates that it is better to recover 3D scene structures based on line segments than that on feature points, at least for some scenes [7][8][9][10][11][12]. For example, for poorly textured scenes, where feature points are hard to be detected and matched, recovering their 3D structures from line matches seems the only choice because their structures can be easily outlined by several edge line segments [13].…”

Image-based reconstruction of wire art

“…While these methods produce impressive results for objects with reliable texture information, they perform poorly when there are insufficient discriminative point features. Some researchers have tackled this limitation by exploring the presence of higher order features such as lines (Baillard et al 1999;Hofer et al 2014Hofer et al , 2013Jain et al 2010) and curves (Xiao and Li 2005). We refer the reader to the recent work of for a theoretical foundation of the multi-view geometry properties of curves.…”

Fast 3D reconstruction method based on UAV photography

“…UAV flight platforms have greatly enhanced the efficiency of data collection [19][20][21]. UAV flight platforms have greatly enhanced the efficiency of data collection [19][20][21].…”

“…Compared with the traditional photogrammetry and remote sensing platform, low-altitude flight UAV and low-cost camera combination meet the 3D image accuracy requirements of digital city construction [18].Using UAV flight formations, many UAVs can be used in a region for data collection. UAV flight platforms have greatly enhanced the efficiency of data collection [19][20][21]. Therefore, the UAV is the best platform for 3D reconstruction of large scenes at this stage.At present, 3D reconstruction results based on multiview images are prone to large voids [22].…”