Multi-view 3D Reconstruction Based on Edge Line Features

Yanxin Zhang

《International Core Journal of Engineering》 6卷8期 (2020/08) Pp. 47-52

https://doi.org/10.6919/ICJE.202008_6(8).0009

摘要

Traditional 3D reconstruction from multiple perspectives is mainly based on the detection and matching of feature points. In the actual target reconstruction scene, environmental light, target surface texture and other factors seriously affect the extraction and accurate matching of feature points, with poor matching effect and low robustness. In addition, the point cloud in the reconstruction process is complicated and requires a lot of computation, so the reconstruction process is time-consuming. Therefore, this paper proposes a method to obtain straight line features based on the edge, and uses this method to obtain straight line features from multi-perspective image sequences, and takes the straight line features as the underlying features to reconstruct the geometric structure of the building scene from multiple perspectives.Through the comparison of experimental results, this algorithm is better than the traditional multiview 3D reconstruction algorithm based on point characteristics.Compared with the existing multi-view 3D reconstruction algorithms based on line features, it is advantageous to establish the position relation between line features and the position relation between regional texture and geometric model, which provides convenience for the texture mapping and mapping of 3D reconstruction geometric model results.

關鍵字

3D reconstruction ； Multi-view ； Line feature ； Edge

參考文獻

Snavely N, Seitz S M, Szeliski R. Photo tourism: exploring photo collections in 3D[C]//ACM transactions on graphics (TOG). ACM, 2006, 25(3): 835-846.

Moulon P, Monasse P , Perrot R , et al. OpenMVG: Open Multiple View Geometry[C]// International Workshop on Reproducible Research in Pattern Recognition. Springer, Cham, 2016.

Wu C. Towards Linear-Time Incremental Structure from Motion[C]// 3DV-Conference, 2013 International Conference on. IEEE Computer Society, 2013:127-134.

Furukawa Y, Curless B, Seitz S M, et al. Towards internet-scale multi-view stereo[C]//2010 IEEE computer society conference on computer vision and pattern recognition. IEEE, 2010: 1434-1441.

Hofer M, Maurer M, Bischof H. Efficient 3D scene abstraction using line segments[J]. Computer Vision and Image Understanding, 2017, 157: 167-178.