Tight-wire Measurement System
(Work in progress...looking for co-conspirators!)
This project is picking up on some preliminary work that I described here on the Open Source Ecology site. There are several useful references there.
A tight wire stretched between two supports can define a very precise straight line in space -- within thousandths of an inch. It has been a common technique for setting up large precise equipment, such as turbine shaft bearings, although it has largely been replaced by laser-based devices.
A wire does sag, of course, under its own weight. However all you need to know to predict the amount of sag precisely is (1) the weight-per-foot of the wire, (2) the length of the free span, and (3) the tension on it. You can provide well-controlled tension by running the wire over a pulley and hanging a weight on it. It is rather cool that with very basic, low-tech components you can measure straightness over 50 or 100 feet within thousandths of an inch!
The old-school way to make measurements relative to the tight wire involved bringing a micrometer head up to the wire and using a continuity tester to determine when you just touched the wire. Tedious! This project uses a low-cost webcam to make the measurement much more convenient.
Wire position measurement (transverse)
I am experimenting with a Logitech C310 webcam which (after a bit of modification) can be manually focused quite close. Here is a view of 0.016-inch-diameter music wire about an inch and a half in front of the camera.
(Reduced from 1280x960 original capture, where scale is approx 0.001 inch per pixel)
Wire position measurement (lengthwise)
A new idea recently added to the tightwire alignment concept is to measure position along the length of the wire. If the wire has markings on it, the camera image can be processed to identify the markings and determine longitudinal position. The idea I am working on is to put black marks on the shiny wire representing a binary code. The binary pattern would be biphase mark encoding representing a maximal-length pseudorandom code of, say, 16 bits. There would be 2^16^-1 unique codes, and by inspecting the image of the wire you could figure out exactly how many steps down the sequence you are looking at. Then, by looking for the exact black/white edge transitions, you could estimate your position down to pixel or even sub-pixel resolution.
Here is a photo of a wire (I used tin-plated Mapes music wire from here) with miscellaneous black-ink markings (not codes!) on it. While I expect lighting and focus to be critical, I consider this to be promising.