Towards Secondary Fingerprint Classification Ishmael S. Msiza, Jaisheel Mistry, and Fulufhelo V. Nelwamondo Biometrics Research Group – Information Security CSIR, Modelling & Digital Sciences Pretoria, Republic of South Africa e-mail: {imsiza,jmistry,fnelwamondo}@csir.co.za Abstract manuscript proposes a move towards the sec- ondary level of fingerprint classification. This is done in order to further penetrate a fingerprint template database, and further reduce it to smaller partitions for efficient execution of the database search procedure. This is done through taking advantage of the extensibility of a structural fingerprint classifier when used on slapped, as opposed to rolled, fingerprints. The said classifier first orders a fingerprint into one of four primary fingerprint classes and, thereafter, into one of eight secondary fingerprint classes. Evaluated on the CSIR-Wits Fingerprint Database (CWFD), the primary classification module registers an accuracy figure of 80.4%, and the secondary classification module registers an accuracy figure of 76.8%. This small difference between the two figures is indicative of the validity of the proposed secondary classification module. Keywords—fingerprint core; fingerprint delta; primary classifi- cation; secondary classification I. I NTRODUCTION The classification of samples in an automated recognition system is primarily important because of the need to virtually divide the template database into smaller, manageable parti- tions. This virtual division is done before executing a database search procedure, and it is done in order to avoid having to search the entire template database and, for this reason, minimize the database search time and improve the overall performance of an automated recognition system. Sample classification is, at a secondary level, important because of its impact on the template database design process. This is because of the fact that, even a good database management system (DBMS) will be negatively affected by a poorly designed database [1]. Even though the concept of sample classification applies to systems that use almost any biometric modality, this manuscript focuses on fingerprint classification, with immediate application to an automated fingerprint recog- nition system. The commonly considered primary fingerprint classes are [2]: Central Twins (CT), Left Loop (LL), Right Loop (RL), Tented Arch (TA), and Plain Arch (PA). Many fingerprint classification practitioners, however, often reduce these five fingerprint classes to four. This is, at a high level, due to the difficulty in differentiating between the TA and the PA class. These two similar classes are often combined into what is referred to as the Arch (A) class. Recent examples of practitioners that have reduced the five-class problem to a four-class problem include Senior [3], Jain and Minut [4], and Yao et al [5]. These four primary classes are normally sufficient in the performance improvement of small-scale applications such as access control systems and attendance registers of small to medium-sized institutions. They, however, may not be suf- ficient in the performance improvement of large-scale appli- cations such as national Automatic Fingerprint Identification Systems (AFIS). In order to enforce visible performance improvement on such large-scale applications, this manuscript introduces a two-stage classification system, by exploiting the extensibility of the classification rules that utilize the locations of the fingerprint global landmarks, known as the singular points [6], on the fingerprint image foreground. The first classification stage produces the primary finger- print classes and then the second classification stage breaks each primary class into a number of secondary classes. It is important to note that the concept of secondary fingerprint classification, for structural fingerprint classifier, is one that has not been exploited by fingerprint classification practitioners. A structural fingerprint classifier is one that uses the arrangement of singular points in order to classify a fingerprint. The next section presents a detailed discussion of both the primary and the secondary fingerprint classes. II. PRIMARY AND SECONDARY FINGERPRINT CLASSES This section presents the proposed primary and secondary fingerprint classes, together with the rules used to determine them. As mentioned before, the rules used to determine these primary and the secondary classes are based on the arrange- ment of the fingerprint singular points, namely, the fingerprint core and the fingerprint delta. Forensically, a fingerprint core is defined as the innermost turning point where the fingerprint ridges form a loop, while the fingerprint delta is defined as the point where these ridges form a triangulating shape [7]. Figure 1 depicts a fingerprint with the core and delta denoted by the circle and the triangle, respectively. A. Central Twins (CT) Primary Class and its Secondary Classes Fingerprints that belong to the CT class are, at a primary level, those that have ridges that either form (i) a circular pattern, or (ii) two loops, in the central area of the print. Some practitioners usually refer to the circular pattern as a whorl [8], while the two-loop pattern is referred to as a twin loop [9]. The similarity, however, between the two patterns is that they both have cores located next to each other in the central area of the fingerprint, which is the main reason why Msiza et al [2] grouped these two patterns into the same class, called This manuscript is a version of a chapter that appears in InTech Publisher’s book on the biometrics series ISBN 978-953-307-488-7. The copyright of any material published by InTech is retained the author 2011 International Conference on Computer Engineering and Applications (ICCEA 2011) V2-249 978-981-08-9196-1/11 ©2011 IACSIT — This