I have just finshed work on version 1 of my Pi powered Panobot. It is a robot designed to take panoramas with the official raspberry pi camera board.
It is built out of lego and uses the same motors and motor controller as my Lego Pibot.
It runs a python script which allows me to program how wide a panorama I want and how many levels. Once the images are taken I copy them over to my laptop and stitch them together into one image using Hugin http://hugin.sourceforge.net/
The Pi camera is very good for its price but compared to my DSLR has a smaller dynamic range and also a smaller HFOV (Horizontal Field Of View) so to get a panorama, it requires a few more pictures. I found to allow enough overlap that 10 pictures wide was sufficient by 3 pictures tall. This totaled to 30 pictures per panorama compared to my normal 25 pictures for my DSLR. I have to stitch them on my laptop instead of the pi due to massive amount of processing power to find the control points
So, time for some pictures


Now for what you have been waiting for, what images does it produce?

For the full sized panobot panorama click here as wordpress limits the upload size

For the full sized DSLR panorama click here as wordpress limits the upload size

For the full sized panorama click here as wordpress limits the upload size

For the full sized panorama click here as wordpress limits the upload size
For the full sized panorama click here as wordpress limits the upload size
As you can see, there really isn’t much in it between the DSLR and the Raspberry Pi Panobot. There is two thing though to take into account though.
Time
The Panobot takes around 2 mins to complete a full panorama
Using the DSLR I took the example panorama in under 30 seconds
Stitching
Because of the larger HFOV on my 28mm lens on my DSLR, I only took 24 photos for the full panorama (In portrait mode to reduce lens distortion). 24 photos takes about 3-4 mins to stitch
The Pi camera board took 30 pictures to get a full panorama. It took 4 mins to stitch those though.
So there is definitely trade off with using a completely automated pi camera but I am still quite pleased with the results. It was a great learning experience and a great option for someone who can’t afford a larger camera, the whole system (excluding lego) runs in about £50-£60
A copy of the python code can be found here
A copy of the python code on github can be found here