Panoramic Image Generation Using Centerline- Constrained Mesh Parameterization for Arbitrarily Shaped Tunnel Lining

Fei Deng,Junxing Yang

doi:10.1109/access.2020.2964121

Abstract

Three-dimensional (3D) triangular meshes has been widely used in tunnel engineering. This paper proposes an algorithm for unwrapping 3D tunnel lining meshes into straight two-dimensional (2D) meshes to generate a 2D seamless panoramic image of the tunnel lining. The proposed algorithm is divided into three steps. In the first step, the centerline is extracted and the L1-median is used to represent the centerline of the tunnel; in the second step, a centerline constraint, which is linear to ensure that the equations to be solved are also linear, is added to least squares conformal mapping (LSCM) to unwrap the tunnel mesh; and in the third step, a panoramic image of the tunnel lining is obtained through texture projection or image correction. Adding a centerline constraint allows the panoramic image to retain mileage information and global shape. The strict geometric relationship between the 3D mesh and panorama guarantees the panorama's accuracy and realizes the two-way mapping of 2D and 3D data analyses. Moreover, a projection strategy is proposed to efficiently process the multiple tiles and non-manifold tunnel meshes which cannot be unwrapped directly using mesh parameterization algorithms. The experimental results show that centerline-constrained LSCM (CLSCM), which is suitable for any arbitrarily shaped tunnel lining, can preserve the global shape and local details better than the conventional method and can be extended to unwrap any strip-shaped structured meshes.

Highlights

The past few decades have witnessed the significant achievement of 3D reconstruction
In this paper, we propose a centerline-constrained mesh parameterization algorithm to unwrap a 3D tunnel mesh into a 2D mesh to generate tunnel lining panorama
1) By adding the centerline constraint, the global shape and local details of the tunnel are better preserved compared with that of the conventional method; Based on the least-squares conformal mapping, centerlineconstrained LSCM (CLSCM) only needs to solve a linear equation system, which greatly improves the efficiency of the algorithm

Summary

Introduction

The past few decades have witnessed the significant achievement of 3D reconstruction. We can obtain high-resolution point cloud data of the object surface using structure from motion(SFM) [1] and dense matching techniques [2]. Surface reconstruction techniques can be used to connect point clouds according to certain topological structures to approximate the surface of objects. Texture reconstruction is performed to map sequenced images to 3D meshes. The 3D meshes have been widely applied to tunnel engineering, such as tunnel inspection [3], building information modeling (BIM) [4], deformation analysis [5], and unwrapping 3D images of rock tunnels [6].

Methods

Results

Conclusion