Mean Squared Error and voting
This is the last step in recognition process.

All histogram members for the first digit are squared and compared to the previously recorded and squared histograms for known numbers (tfis process is not described here). All differences are added and the table of 10 sums is created. Then the decision is made which one is the winner. Also, the next best match is found, and its rating is calculated, using formula  ((best result) / (second best result) - 1) * 100. The result shows the ratio between first and second best number in %.

If it is 2-digit number, the same process is made with the second digit. Now the number is recognized, and the worse digit rating is used as the total recognition rating. In this case, the worse digit (we don't know which one) has the second best match with 147% differences more than the first match (that means 2,47 times worse result).



    Back to projects 
0. Home
H. Hardware
L. Lighting
1. Frame fetch
2. Ball locate
3. Stretching
4. Unsharp mask
5. Component selection
6. Holes and scratches
7. Component shrink
8. Invert selection
9. Component list
10. Select best circle
11. Angle detection
12. Rotate selected
13. Select broad lines
14. Selected components list
15. Components scaling
16. Histogram
17. MSE and voting
C. Conclusion