3D Reconstruction of Weak Texture Image Based on Tensor Voting and Dense Map Method

Abstract

Abstract: Image based 3D reconstruction method can't effectively detect and match feature points and can't robustly expand 3D point cloud in weak texture image region, due to the fact that weak texture has single color and local reflection phenomenon. It is possible to produce model hole and a large number of 3D noise points, and affect the accuracy and integrity of 3D reconstruction. Slack variables constraint was used to filter mismatch features optimizing the camera matrix, in the dense point cloud expansion stage by using tensor voting principle to filter noise 3d point which is inconsistent with three-dimensional point's normal around. Multi-scale discrete-continuous variational method was used to reconstruct 3D models. The experimental results show that the proposed method has better accuracy and integrity than the reconstruction methods of PMVS, MVE and MeshRecon.

Key words: weak texture, tensor voting, multi-scale discrete-continuous, 3D reconstruction

CLC Number:

TP391.9

Chen Mingwei, Yang Yang. 3D Reconstruction of Weak Texture Image Based on Tensor Voting and Dense Map Method[J]. Journal of System Simulation, 2015, 27(10): 2553-2559.

References

[1] N Snavely, S M Seitz, R Szeliski.Photo tourism: Exploring photo collections in 3D[J]. ACM Transactions on Graphics (S0730-0301), 2006, 3(25): 835-846.
[2] Yasutaka Furukawa, Brian Curless, Steven M Seitz.Towards Internet-scale multi-view stereo[C]// IEEE Computer Vision and Pattern Recognition. USA: IEEE, 2010: 1434-1441.
[3] M Kazhdan, H Hoppe.Screened poisson surface reconstruction[C]// ACM Transactions on Graphics. USA: IEEE, 2013, 32(3): 1-13.
[4] Byungjoon Chang, Sangkyu Woo, Insung Ihm.GPU-based parallel construction of compact visual hull meshes[J]. The Visual Computer (S0178-2789), 2014, 30(2): 201-211.
[5] Sorensen, Rhein, Mahoney. Reconstruction of textureless regions using structure from motion and image-based interpolation[C]// International Conference on Image Processing. USA: IEEE, 2014: 1847-1851.
[6] 陈明伟, 杨洋. 弱纹理图像建模中的相机优化和扩展滤波研究[J]. 系统仿真学报, 2013, 25(10): 2449-2453.
[7] 缪君, 储珺, 张桂梅. 基于稀疏点云的多平面场景稠密重建[J]. 自动化学报. 2014, 41(4): 813-822.
[8] S Fuhrmann, F Langguth, M Goesele.MVE A Multi View Reconstruction Environment[C]// Euro graphics Workshop on Graphics and Cultural Heritage. Germany: Eurographics, 2014.
[9] Zhuoliang Kang, Gerard Medioni.Fast Dense 3D Reconstruction using an Adaptive Multi-scale Discrete-Continuous Variational method[C]// Winter Conference on Applications of Computer Vision. USA: IEEE, 2014: 53-60.
[10] Changchang Wu, Sameer Agarwal, Brian Curless, et al. Multicore Bundle Adjustment[C]// Computer Vision and Pattern Recognition. USA: IEEE, 2011: 3057-3064.
[11] Carl Olsson, Olof Enqvist.Stable Structure from Motion for Unordered Image Collections[C]// Conference on Image Analysis. USA: IEEE, 2011: 524-535.
[12] T Wu, S Yeung, J Jia, et al. A Closed-Form Solution to Tensor Voting: Theory and Applications[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence (S0162-8828), 2012, 34(8): 1482-1495.
[13] A Chambolle, T Pock.A first-order primal-dual algorithm for convex problems with applications to imaging[J]. Journal of Mathematical Imaging and Vision (S0924-9907), 2011, 40(1): 120-145.
[10] W E Lorensen, H E Cline.Marching cubes: A high resolution 3d surface construction algorithm[C]// ACM Siggraph Computer Graphics. USA: IEEE, 1987, 21: 163-169.