I am working in semiconductor inspection and the tools we have are tightly controlled for temperature.
Seems to me that the variation you are looking at can come from ambient temperature.
Temperature changes are usually slow process and can be calibrated periodically.
One solution is like suggested before by Bruce Amon is to have calibration scale next to the object.
The other solution is to take intermediate image of calibration pattern between measurements and recalibrate your optical system.
There is recommended calibration grid at: "C:\Users\Public\Documents\National Instruments\Vision\Documentation\CalibrationGrid.pdf"
You can calibrate your optical system between measurements.
See vision concepts help how to use that.
Amit
Amit Shachaf
