Target-less computer vision for traffic signal structure vibration studies

•Target-less computer vision is developed for structural dynamic response studies.•Dynamic properties determined with consumer-grade camera and sub-pixel vibrations.•Mode shapes accurately determined with maximum RMS modal displacements of 0.5 pixel.•Wind-induced equivalent stress ranges calculated and validated with high accuracy.

The presented computer vision method allows for non-contact, target-less determination of traffic signal structure displacement and modal parameters, including mode shapes. By using an analytical model to relate structural displacement to stress, it is shown possible to utilize a rapid set-up and take-down computer vision-based system to infer structural stresses to a high degree of precision. Using this computer vision method, natural frequencies of the structure are determined with accuracy similar to strain gage and string potentiometer instrumentation. Even with structural displacements measured at less than 0.5 pixel, excellent mode shape results are obtained. Finally, one-minute equivalent stress ranges from ambient wind excitation are found to have excellent agreement between the inferred stress from strain gage data and stresses calculated from computer vision tied to an analytical stress model. This demonstrates the ability of this method and implemented system to develop fatigue life estimates using wind velocity data and modest technical means.