I would have thought a simple device that counts the number of times the wheel has rotated would be vastly simpler and cheaper, and surely accurate to within a few centimeters.
But fiberoptic gyro navigation is already a thing, and has been for decades. It's not super cheap, but it's cheap enough for an industrial application like this one. So I'm not sure what problem is being solved here.
* What would the Central Limit Theorem give us if a significant number of wheels on loco's and carriages were instrumented and ID logged over the years.
* What drift patterns and correction factors can be observed and fine tuned by passing {many} wheel streams through station to station and line end to line end "known distance" corrections.
For half a century or so aircraft height was largely done with air pressure alone.
To this day GPS accuracy is "improved" by logging the "drift" of fixed point stations giving us filter coefficients to correct for atmospheric wobble, acute angle uncertainty, as yet uncorrected time drift, etc.
It still is. Flight levels (20000 feet and above in the US) are defined by air pressure without reference to current atmospheric conditions (i.e. the actual altitude of FL200 can vary based on atmospheric conditions). Below 19000 feet, altimeters are calibrated to local conditions which are given to the pilots by air traffic control.
See: https://en.wikipedia.org/wiki/Altimeter that list five types (including trad. air pressure type)
Also trains have quite a large number of axles, I'm sure you could reduce error by counting all of them.
A curious advantage of instrumenting all axles for wheel turn counting is calibration against actual known distances (end to end, etc) over time reveals slow wheel degradation that can trigger carriage maintainance.
Not all operators do this, it's been done at various times for heavy rail mining operators with loooong trains and heavy daily usage.
https://www.universalsignalling.com/news/latest#h.12nuchpah7...
The spit up of grease, leaves, metal particulate from rail and wheel grind, etc gets pretty thick.
The vision challenge is to cheaply and effectively keep a clean lens.
There's a similar challenge in monitoring production conveyor belts in mineral processing - vibration, tempreture, rocks kick about, dust make for a savage environment.