UAS-based bridge displacement measurement using two cameras with non-overlapping fields of view

Traditionally, when a bridge fails an assessment, information required for bridge load capacity may be obtained by fitting a bridge with sensors and measuring the bridge's response to a load of known weight being driven across. Unfortunately, fixing sensors to the bridge is cumbersome and time-consuming, requiring erection of scaffolds and platforms and so can be expensive. Camera-based measurements using Unmanned Aerial System (UAS) promise an alternative non-contact approach that is safe, affordable, and fast. However, the lack of stationary locations near bridge mid-span makes it difficult to both measure displacements and stabilise the ego motion of a UAS. To overcome this challenge, this paper therefore presents the experimental validation of a method that uses two different cameras mounted on a UAS. One camera is focused on taking measurements near the midspan while another camera is used to stabilise the UAS using stationary locations at the bridge supports. The method does not require that the stationary object is in the field of view of the measurement camera. The method only uses video feed from two cameras viewing different parts of the bridge and no additional sensor information. The significance of this paper to the field of UAS bridge displacement measurements is that the method opens up the ability to measure displacements in locations such as midspans where stationary objects are rare. Submillimetre accuracy was achieved indoors on a model bridge setup and outdoors on a 5 m bridge analogue. • A new UAS bridge displacement measurement technique using two different cameras with non-overlapping fields of view. • Method exploits the Euclidean homography from one camera to correct for ego motion in the displacement measurement camera. • The proposed method is validated both indoors and outdoors. • The method has potential to achieve simultaneous vehicle tracking and bridge displacement measurement.