We can provide near unlimited samples of characters to train tesseract on.
they are all digits 0-9. Each image is a single character, so it is easy to group them into 0's, 1's, 2's etc.
The images are of high enough quality with some basic training we should expect very accurate OCR.
This is an urgent task and we can discuss milestones based on quick delivery.