초록 ▼

An image comparison arrangement uses an electronic computer to compare digitized fingerprint minutia maps of fingerprints of an unknown fingerprint set with corresponding maps of reference fingerprint sets which are stored in memory, in order to identify unknown fingerprints or to match fingerprints

An image comparison arrangement uses an electronic computer to compare digitized fingerprint minutia maps of fingerprints of an unknown fingerprint set with corresponding maps of reference fingerprint sets which are stored in memory, in order to identify unknown fingerprints or to match fingerprints. The matching is performed by converting all the fingerprints to attributed relation graphs (ARGs) including nodes and branches, to which attributes are appended. For each fingerprint pair being compared, a distance matrix is generated, the elements of which are the similarities of stars. The highest-ranking star pair is selected as the starting point of a comparison tree, by which attempts are made to fill a match core with elements representing the matching stars. The comparison is of the various attributes of the nodes and branches of each star. Once the maximum consistent number of stars has been matched in each fingerprint set, the next reference fingerprint is compared with the unknown fingerprint, until all relevant reference fingerprints have been compared. The number of elements in the match core indicates the degree of match of the unknown to each reference fingerprint.

대표청구항 ▼

A method for matching a set of unidentified fingerprints, which set includes at least one unidentified fingerprint, with a plurality of sets of reference fingerprints from a fingerprint file, said method comprising the steps of: generating an attributed relational graph (ARG) including (a) nodes and

A method for matching a set of unidentified fingerprints, which set includes at least one unidentified fingerprint, with a plurality of sets of reference fingerprints from a fingerprint file, said method comprising the steps of: generating an attributed relational graph (ARG) including (a) nodes and node attributes and (b) branches between said nodes, and branch attributes, from an extracted digital minutia map of said set of unidentified fingerprints, to thereby implicitly generate stars centered at each of said nodes; generating a distance matrix between (a) the stars in said ARG of one of said fingerprints of said set of unknown fingerprints and (b) the stars of the ARG of one of the fingerprints in one of said sets of reference fingerprints, said distance matrix including a matrix element associated with each pair of stars; sorting said elements of said distance matrix for each fingerprint pair, according to the value of said elements, to form a sorted distance matrix which establishes an order of star pair matches; generating a match core of consistent sets of star pairs from said distance matrix and said ARG in an order established by said sorted distance matrix; wherein said step of generating a match core includes the steps of: selecting a star pair associated with a highest element of said sorted distance matrix, to define a first element of a match core; deleting from said distance matrix that element associated with said first element of said match core, to generate a reduced distance matrix; selecting, from among candidate pairs of stars centered on neighbor nodes of the center nodes of said star pair in said first element of said match core, that candidate star pair which is both (a) consistent with said match core, and (b) among all said candidate star pairs which are consistent with said match core, associated with a highest distance matrix element in said reduced distance matrix, to thereby generate a second element for said match core; adding said second element to said match core as a further element; deleting from said reduced distance matrix that one of said elements of said distance matrix associated with said candidate star pair added to said match core, to form a further reduced distance matrix; repeating said steps of selecting that further star pair, adding, and deleting from said reduced distance matrix, at least until no more candidate star pairs consistent with the match core remain.