Fingerprint Classification Based on Maximum Variation in Local Orientation Field Shekhar Suralkar Electronics dept. SSBTE, Bambhori Jalgaon , India Milind E Rane, Pradeep M. Patil Electronics dept. Vishwakarma Institute of technology, Pune, India Patil_pm@rediffmail.com Abstract— Fingerprint classification provides an important indexing mechanism in a fingerprint database. Accurate and consistent classification can greatly reduce fingerprint-matching time and computational complexity for a large database as the input fingerprint needs to be matched only with a subset of the fingerprint database. Classification into six major categories (whorl, right loop, left loop, twin loop, arch, and tented arch) with no reject options yields an accuracy of 89.7 %. The overall accuracy is improved to 91.5 % if the arch and tented arch are merged as a single class. The penetration rate of the proposed classification system is 88.9%. Keywords— Fingerprint classification, whorl, right loop, left loop, twin loop, arch, orientation field, singularity points I. INTRODUCTION The identification of a person requires a comparison of his fingerprint with all the fingerprints in a database. This database may be very large (e.g., several million fingerprints) as in many forensic and civilian applications. In such cases, the identification typically has an unacceptably long response time. The identification process can be speeded up by reducing the number of comparisons that are required to be performed. A common strategy to achieve this is to divide the fingerprint database into a number of bins (based on some predefined classes). A fingerprint to be identified is then required to be compared only to the fingerprints in a single bin of the database based on its class. Fingerprint classification refers to the problem of assigning a fingerprint to a class in a consistent and reliable manner. Although fingerprint matching is usually performed according to local features (e.g., minutiae), fingerprint classification is generally based on global features, such as global ridge structure and singularities. Fingerprint classification is a difficult pattern recognition problem due to the small inter-class variability and the large intra-class variability in the fingerprint patterns Moreover, fingerprint images often contain noise, which makes the classification task even more difficult. The selectivity of classification-based techniques strongly depends on the number of classes and the natural distribution of fingerprints in these classes. Most of the existing fingerprint classification methods can be coarsely assigned to one of these categories: rule-based, syntactic, structural, statistical, neural network- based and multi-classifier approaches. A fingerprint can be simply classified according to the number and the position of the Singularities. In [1], the Poincare index is exploited to find type and position of the singular points and a coarse classification is derived. The problem with this method is that fingerprints of the Arch type do not have any singularity in terms of Pointcare Index so structural heuristic is used to locate the core point. A syntactic method describes patterns by means of terminal symbols and production rules; a grammar is defined for each class and a parsing process is responsible for classifying each new pattern [2, 3]. Structural approaches are based on the relational organization of low- level features into higher-level structures. This relational organization is represented by means of symbolic data structures, such as trees and graphs, which allow a hierarchical organization of the information [4]. These regions and the relations among them contain information useful for classification. In statistical approaches, a fixed-size numerical feature vector is derived from each fingerprint and a general- purpose statistical classifier is used for the classification [5]. Many approaches directly use the orientation image as a feature vector, by simply nesting its rows [6, 7]. Most of the proposed neural network approaches are based on multilayer perceptrons and use the elements of the orientation image as input features [8]. A pyramidal architecture [9,10,11] constituted of several multilayer perceptrons, each of which is trained to recognize fingerprints belonging to a different class. Neural networks can be extensively used for fingerprint classification [12]. Fingerprints are classified as Lasso or Wirbel [13]. In this paper we introduce a process for classification of fingerprints based on maximum variation in local orientation field. The paper has been organized as follows. Section II introduces the proposed algorithm. Section III explains the required parameters for judging the performance of fingerprint classification. Section IV displays the results and a brief discussion is carried out based on these results. II. PROPOSED ALGORITHM The proposed algorithm is divided in two steps. The first step computes the singularity points on the fingerprint image based on the maximum variation of its local orientation. The second step classifies the fingerprint based on the location of the detected core and delta points. Proceedings of the 2009 IEEE International Conference on Systems, Man, and Cybernetics San Antonio, TX, USA - October 2009 978-1-4244-2794-9/09/$25.00 ©2009 IEEE 945