The first solution that comes to mind, is to simply take a picture, either with wide-angle from close distances or zoomed-in from far away. Personally I don't believe that it was done either way, because if shot from far away other objects would have obstructed the picture. If shot in wide-angle the distortions and viewing-angle at the edges would be way bigger.
One idea that came to mind to create such pictures would be to mount a one-dimensionale CCD sensor on top of a moving object that will be dragged along the street while recording. It might also be possible to use a standard filming-camera. The challenge with that would be to extract a single column of pixels out of each frame shot. It might also be beneficiary to film vertical to increment the resolution of the column. See sketch below.
At the top you see the movie-reel segmented in frames. The center column of each frame should then be extracted and put together to create the new single image you see at the bottom.
As creating a proper setup would be quite expensive I try to approach that idea slightly different. Proud owner of a GoPro Hero3 it was obvious for me to utilize the things I have at hand. Having found a substitution for the CCD sensor, the next step is to define a proper moving object, like a car. And thats all there is for the setup. Another sketch illustrates the plan.