Measuring bridge vibrations and deflections with a camera [improved]

In my previous post I was 1.7 km away and I attempted to measure the deflections of bridge 25 de Abril in Lisbon using OpenCV in Python. This time, I got closer to the bridge and filmed it again. I had to use a tripod this time, but unfortunately it was a windy day and the camera vibrated quite a lot.

In a first attempt, I used the same code as in my previous post to see what happens. The result is shown in the figure below. The last big "wave" in the figure is due to the passage of a train. Not bad, considering that the camera itself vibrated significantly. However, the gradual decrease of the ordinate of the tracked point is not realistic and it is likely due to movements of the camera (besides wind, I used a cheap tripod).

Clearly, it is necessary to improve the code in order to take into account the movement and vibrations of the camera. To this end, I added a tracking point located in a building in Lisbon (which can be assumed to be fixed):

I tracked both points for each frame of the video, and I calculated the changes in the relative distance between these two points. Here is the final result:

Ideally, a comparison with accelerometers or displacement transducers installed on the bridge would be the next logical step to check whether the vibrations in the figure above are due to traffic plus wind or due to errors and limitations of the code. Again, the passage of the train is clear (after time 300 seconds until the end). It is even clearer in the new video (this time I included real-time plotting):

If you are interested in learning how to code the procedure I followed to measure bridge vibrations, please read my new blog post.

If you liked this post or if you think someone you know might benefit from it, please share and comment down below. Stay tuned for more!

...