Fingerprint Classification Using Orientation Field Flow Curves

Manual fingerprint classification proceeds by carefully inspecting the geometric characteristics of major ridge curves in a fingerprint image. We propose an automatic approach of identifying the geometric characteristics of ridges based on curves generated by the orientation field called orientation field flow curves (OFFCs). The geometric characteristics of OFFCs are analyzed by studying the isometric maps of tangent planes as a point traverses along the curve from one end to the other. The path traced by the isometric map consists of several important features such as sign change points and locations as well as values of local extremas, that uniquely identify the inherent geometric characteristics of each OFFC. Moreover, these features are invariant under changes of location, rotation and scaling of the fingerprint. We have applied our procedure on the NIST4 database consisting of 4,000 fingerprint images without any training. Classification into four major fingerprint classes (arch, left-loop, right-loop and whorl) with no reject options yields an accuracy of 94.4.%