How does the system handle edge cases (such as; differentiating between 0 & O, 5 & S,8 & B, X& K)?
- Keypoint-Based Features with SIFT Descriptor Matching: The SIFT algorithm extracts robust features that uniquely represent characters. These features can capture specific details that distinguish “0” from “O” or “5” from “S”.
- Euclidean Distance: The system computes the Euclidean distance between test descriptors and the reference database. The closest match determines the recognized characters.
- Voting Mechanism(k-neighbour) for Ambiguity: A voting system is applied when the Euclidean difference is not high enough. This ensures that the most frequently occurring character is selected, reducing misclassifications.
- Reference Database: The reference database with accurately labeled examples of all characters ensures better recognition. Our system uses slight variations(rotated versions, low-quality versions) in character appearance to make distinctions.
