Nontarget Vision Sensor for Remote Measurement of Bridge Dynamic Response

Abstract

Displacements of railroad bridges under trainloads need to be closely monitored, but conventional displacement sensors have limitations for use in the field. This paper presents a new vision-based sensor system developed for remote measurement of structural dynamic displacements without requiring a specially installed target-marker panel. By implementing a robust object-search algorithm, the displacement can be accurately measured by tracking existing bridge surface features from a remote distance. The accuracy of measured dynamic displacements was first evaluated using a shaking table test. Then field tests were carried out on two railroad bridges subjected to freight trainloads traveling at various speeds. Measurements were taken remotely during the daytime and also at night from different distances with and without a target panel. Through comparison with a conventional contact-type displacement sensor, the high accuracy of the proposed nontarget remote-sensor system was demonstrated in the realistic field environments. From the measured displacement time histories, frequency-domain characteristics associated with the train–bridge systems were further analyzed, confirming the capability of the vision system in measuring high-frequency components. By targeting existing features on a structure without requiring a target panel installed on a fixed location of the structure, the vision sensor system developed in this study provides the flexibility to easily change displacement measurement locations, in addition to other advantages, such as easy setup and no need to access the structure.