Geometry based hand vein biometry

This thesis documents a study in which new algorithms are developed for geometrybased hand vein biometry. Hand vein patterns are assumed not to change overtime except in their size, and they are unique to each individual, hence researchers aimto construct a biometric control system based on hand vein patterns. The approachproposed here is using free-posture captured near infrared hand vein images for bothenrollment and test. We utilize Gabor filters banks to enhance the visibility of handvein segments which is a new thing in hand vein biometry literature. Furthermore, it isrobust against wrongly aligned hand vein features. The region of interests (ROIs) areextracted from hand vein images. In order to increase the visibility of hand veins inthese ROIs, Gabor filter bank approach is applied. Enhanced ROIs facilitate to extractvein line segments as geometric features. To extract these line segments; thresholding,skeletonization and line edge map (LEM) extraction methods are applied on enhancedROIs, respectively. These methods yield the LEM and the graph version of a hand veinstructure. Before identity matching, a preprocessing stage is configured to alleviate theefects of wrong registrations. There are two different methods for alignment correction.The first one is based on keypoint matching, whereas the other is based on translatingthe compared LEMs iteratively. Last but not least, identity matching is done by severaldistance measurement metrics, namely, line segment Hausdor distance (LHD),weighted line segment Hausdorff distance (WLHD), modified line segment Hausdorffdistance (MLHD) and graph edit distance (GED). WLHD and MLHD are different versionsof LHD, that we specialized for hand vein biometry. Additionally, an algorithmis developed to make a measurement on Graph Edit Instance (GED) metric. GED isdefined as the least cost graph edit operation sequence which is used to transform onegraph to another. By examining each metric, we notice that line segment matchingbased methods give more promising results than graph matching.