Tuesday, March 20, 2018

Photogrammetry of the FIU Bridge Collapse (partial)

This is a partial photogrammetric reconstruction of the FIU bridge collapse.




Photogrammetry is the science of measuring the spatial change between two similar reference photos. In the use for this video, each real frame from the (original?) source was mapped with reference points and then motion was interpolated based on the change in the position of the dots.




Without some extra processing, photogrammetry has it's drawbacks. 

One error you may notice in the video above is several rapid shifts in the motion and possibly an apparent increase in the speed of the motion as well. This is an aberration of the motion switching between the real frames of the original video. If an object is following an arcing path in its motion, drawing two reference dots will only give you the linear path between the two unless corrections in this motion path are made. I've not made these corrections in this because it can be challenging to get a proper motion path in an event that has numerous independent moving objects that each have their own likely non-linear path. I found it best to just iterate linear paths between the frames and your own judgement can fill in the true path of the objects.

Another quite apparent aberration is the movement of objects when they cross the path of another object. The process of photogrammetry only measures the distance between two reference dots to discern motion. If that series of dots intersects another set of dots while transposing their motion, it can create some quite funky image tearing. Again, this is a correctable issue using masking techniques to separate the objects followed by recompositing everything back together. This can become quite time intensive so I have not performed these additional corrections.

 In the set of frames I grabbed from the source I used, I captured 30 frames. The video above only takes you from frame 1 to frame 17. An additional 4 frames are needed for about 98% of the collapse to have taken place. I have not yet mapped these frames and I'm a bit short on time to do this now. If there is enough interest, I'll make the time however.

I do apologize for not attributing the person who originally contributed this video. I am not aware who made it available first. The highest quality source I found on Youtube was from a user going by the name
OfficialJoelF. I don't know if this is the originator for the video or not though. Thank you for whoever was the driver of this truck for providing it to the public. It is such an incredibly lucky video in so many respects and will likely be quite useful in determining the full scope of the issues this bridge experienced in its short time.