Reference-free ego-motion compensation for vision-based measurement of structural deformations of miter gates

Abstract

Monitoring the structural condition of miter gates in locks and dams on inland waterways is vital to prevent unforeseen failures. Considering that the application of hydrostatic loads on miter gates is a quasistatic process resulting in two primary structural states, monitoring displacements induced by these loads is an effective method for indicating common degradations and examining miter gate safety. Noncontact sensors such as dual-camera systems are preferred for ensuring continuous operation. A dual-camera system records pre- and postload three-dimensional positions of a miter gate, allowing for the calculation of load-induced structural displacements. However, movement of the dual-camera system itself during the measurement process, referred to as “ego-motion,” introduces measurement errors, as if the structure undergoes additional rigid body motion. Existing studies utilize stationary reference points to correct false displacements; however, the reference points are often challenging to identify in the field. Instead of displacement, this article introduces an algorithmic approach for vision-based extraction of structural deformation while mitigating the impact of rigid body motion from both the structure and the dual-camera system. The algorithm is formulated on the premise that the measurement of deformation, which indicates changes in the relative positions of points on a structure, remains unaffected by rigid body motions originating from any sources. Besides the measurement robustness under ego-motion, deformation is often more informative than displacement in detecting degradations in structures by providing a more direct link to structural strain. The proposed approach is validated through numerical simulations focusing on a field measurement scenario at a miter gate located at the Dalles, Oregon. The procedure of deformation measurement utilizing a dual-camera system is also demonstrated in the synthetic environment created by a graphics-based digital twin of the gate. The proposed approach will facilitate inexpensive deformation measurements that enable condition assessment of miter gates.