|
Title |
Probabilistic 3D Position and Displacement Estimation of Measurement Points on Civil Structures Based on Stereo Vision
|
|
Authors |
이가민(Gamin Lee) ; 이준화(Junhwa Lee) |
|
DOI |
https://doi.org/10.11112/jksmi.2026.30.4.26 |
|
Keywords |
스테레오 비전; 3차원 변위 계측; 불확실성 정량화; 오차 전파; 공분산 타원체; 토탈스테이션 Stereo vision; 3D displacement measurement; Uncertainty quantification; Error propagation; Covariance ellipsoid; Total station |
|
Abstract |
This study proposes a probabilistic stereo-vision framework that propagates matching uncertainty to 3D position covariance and displacement covariance for civil-structure monitoring points. Per-point pixel covariance is assigned from the curvature of the zero-mean normalized cross-correlation (ZNCC) peak, and the Jacobian of closest-ray midpoint reconstruction is used for first-order error propagation. The reconstructed point is expressed as and the displacement as , visualized as a covariance ellipsoid. Simulations confirm agreement between the analytical and Monte Carlo covariances. A field experiment on a pedestrian suspension footbridge visualizes the tracked cable-intersection points, their covariance ellipsoids, and representative displacement traces, showing how the resolvability differs between near and mid-range points. The framework therefore reports not only displacement estimates but also point-wise measurement reliability.
|