Long Deck Suspension Bridge Monitoring: The Vision System Calibration Problem

Abstract

Structural Health Monitoring is an emergent powerful diagnostic tool that can be used to identify and to prevent possible failures of the various components that comprise an infrastructure. In the particular case of a suspension bridge, the measurement of the vertical and transversal displacements plays an important role for its safety evaluation. Taking into account the restrictions usually found on these structures, an enhanced solution comprises a non‐contact vision‐based measurement system with dynamic response, accuracy and amplitude range well suited to the physical phenomenon to measure. The paper describes a methodology to perform the vision system calibration that can be carried out in‐situ, while the deck is moving, requiring little effort and a minimum set of information. Specifically, only a set of active targets fixed on the deck and the knowledge about the distances between them is required. Results related to the performance evaluation, obtained by numerical simulation and by real experiments with a reduced structure model, are presented and they show that, even in an environment severely affected by disturbance noise, it is possible to measure the vertical and transversal displacements with a standard accuracy better than 10 mm.

Simultaneous vision system calibration and full-motion estimation using a sequence of noisy images from a stereo affine cameras

“… To improve the algorithm's convergence at the start off, a gross initial value of the cameras' projection matrices (M) and of the object (L) is required. To fulfil this goal, we apply the factorization method to the first image followed by a scaling, based in the knowledge of the distance between the points [1].…”

“…A common solution is to measure the acceleration or the velocity and integrate the measured values in time domain. While in principle it is possible to recover the time history of the bridge deck displacement, it is well known that this method has several drawbacks [1,2]. An enhanced solution comprises a non-contact measuring system with dynamic response, accuracy and amplitude range well-suited to the physical phenomenon to measure [1].…”

“…Amongst the set of non-contact measurement systems the vision-based systems are the more advantageous solution for tracking the motion of large structures, namely those built with steel material [1]. However, the number of vision-based measurement systems reported on the literature is very limited and most of them use very simple camera and motion models, which quickly fail when the structure displacement occurs in a plane not parallel to the image plane [2,5,6,7,8,9,10].…”

“…Though, to obtain measurements with the aforementioned accuracy an adequate vision system calibration and motion estimation procedures are required. Usually, the vision system calibration and the object's motion estimation procedures are performed independently, in separated steps, with the calibration being carried out in offline mode [1]. In this paper we will present a recursive algorithm that allows performing the vision system calibration and, simultaneously, estimate the full motion of the structure over time.…”

“…By now, only a few measurement systems satisfy these requirements, albeit partially. This set includes the systems based on: i) GPS (Global Position System) [3]; ii) radar technology [4]; and iii) optical devices [1,2,5,6,7,8,9,10].…”