I was testig with OCR but do not run.
OCR depends on a good quality image to recognize known Characters.
It can be of much help if you first crop the part of the image with the Odometer. Run the OCR on the Cropped image then.
I mean the red frame in the exampleimages.
Best result you´ll get with exact frame size and position though.
I did something similar with the results from a Playstationgame.
The image is converted to grayscale to rise the good results first.
The Image is then splittet into smaller ones.
These images are recognized through Tesseract then.
See the attached zip for the single images and the tesseract result files for them.