Paper Info
Title
Geometrically Driven Underground Camera Modeling And Calibration With Coplanarity Constraints For A Boom-Type Roadheader
Abstract
The conventional calibration methods based on a perspective camera model are not suitable for the underground camera with two-layer glasses, which is specially designed for explosion proof and dust removal in a coal mine. Underground camera modeling and calibration algorithms are urgently needed to improve the precision and reliability of underground visual measurement systems. This article presents a novel geometrically driven underground camera calibration algorithm for a boom-type roadheader. The underground camera model is established under coplanarity constraints, considering explicitly the impact of refraction triggered by the two-layer glasses and deriving the geometrical relationship of equivalent collinearity equations. On this basis, we perform parameters calibration based on a geometrically driven calibration model, which are 2D-2D correspondences between the image points and object coordinates of the planar target. A hybrid Levenberg-Marqurdt (LM) and particle swarm optimization (PSO) algorithm is further proposed in terms of the dynamic combination of the LM and PSO, which optimizes the underground camera calibration results by minimizing the error of the nonlinear underground camera model. The experimental results demonstrate that the pose errors caused by the two-layer glass refraction are well corrected by the proposed method. The accuracy of the cutting-head pose estimation has increased by 55.73%, meeting the requirements of underground excavations.
Year
DOI
Venue
2021
10.1109/TIE.2020.3018072
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
Keywords
DocType
Volume
Cameras, Calibration, Glass, Mathematical model, Pose estimation, Coal mining, Camera calibration, nonlinear optimization, two-layer glasses, underground camera model, vision-based pose estimation
Journal
68
Issue
ISSN
Citations 
9
0278-0046
0
PageRank 
References 
Authors
0.34
0
4
Name
Order
Citations
PageRank
Wenjuan Yang102.70
Xuhui Zhang254.31
Hongwei Ma33610.57
Guang-Ming Zhang452.00