This publication describes a random sample consensus algorithm with enhanced latency for fingerprint matching. A fingerprint image can be represented by a fingerprint image template which includes key points (X and Y locations) and feature vectors. Key points correspond to various minutiae points of the fingerprint image. Feature vectors are rotationally invariant encodings of image blocks centered around key points. Fingerprint matching is done by comparing feature vectors from a verify fingerprint image template to feature vectors from an enrolled fingerprint image template to generate matching key point pairs. A geometric transformation between the verify fingerprint image template and the enrolled fingerprint image template is inferred by a random sample and consensus process of matching key point pairs. The geometric transformation between two matching fingerprint image templates is mainly a rigid two-dimensional (2D) transformation with a translation vector, a rotation matrix, and minimal stretch. Rather than comparing every matching key point pair to infer the geometric transformation, the disclosed fingerprint matching algorithm implements a stretch ratio check. For any two key points, the stretch ratio is the distance between the two key points from the verify fingerprint image template divided by the distance between matching key points from the enrolled fingerprint image template. If the stretch ratio falls outside an acceptable range of stretch ratios, the fingerprint matching algorithm skips the set of matching key point pairs of that stretch ratio when inferring the geometric transformation. By so doing, the fingerprint matching algorithm improves matching speed and makes matching and non-matching latencies consistent.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.