Article Preview
Top1. Introduction
Biometric human identification is based on the features extracted from digital images. Classification is one of the most important tasks to be done during the image processing (Aziz et al., 2013a,b, 2012; Inbarani et al., 2016; Hassanien et al., 2019a,b, 2017, 2014a,b, 2015). Digital image classification is the process of sorting all the pixels in an image into a finite number of individual classes. Ear classification points out the method by which an input ear image is assigned to one of many predefined classes based on a set of features extracted from the images (Pflug et al., 2014). Feature Selection (FS) is an important part of knowledge discovery to improve the classification accuracy and reduce the computational time of classification algorithms (Mitra et al., 2002; Pedergnana et al., 2012; Chandrasekhar et al., 2012; Asad et al., 2012, 2014a,b,c,d; Jothi et al. 2013; Azar et al., 2013a,b; Inbarani et al., 2012, 2013, 2015a,b; Inbarani and Banu, 2012; Zhu and Azar, 2015; Sayed et al., 2019, 2020). In terms of feature selection methods, they fall into the filter and wrapper categories. In filter model, features are evaluated based on the general characteristics of the data without relying on any mining algorithms. On the contrary, wrapper model requires one mining algorithm and utilizes its performance to determine the goodness of feature sets (Hassanien et al., 2015). The selection of relevant features is important in both cases (Banu et al., 2014). Hence a rough set-based feature selection method is applied to the proposed work.
In 1980’s, Zdzislaw Pawlak introduced a mathematical tool called rough set theory (Pawlak, 1982). The basic idea behind the rough set theory is to handle uncertainty and vagueness in some datasets. Pawlak rough set model is constructed using equivalence class and equivalence relations (Pawlak, 1984, 2002). Rough set theory can handle various challenges like segmentation, clustering, feature selection and pattern recognition (Pawlak, 1982, 1993, 2002, 2012; Wang et al., 2007; Thangavel and Pethalakshmi, 2009; Velayutham and Thangavel, 2011; Azar et al., 2015, 2016, 2017; Inbarani et al., 2014a,b,c, 2016; Jothi et al. 2017). These challenges are analyzed using various image modalities like remote sensing image, CT image, biometric images etc. The neighborhood rough set approach is based on finding neighborhood pixels in an image. This paper presents the application of neighborhood rough set classification for person identification using ear images. The emerging field biometric technology for analyzing a person is based on physical and behavioral characteristics. These characteristics are unique and vary from person to person for verification and identification. Biometrics identification techniques are proved to be very efficient and easy for users than conventional methods of human identity. Biometrics techniques are applied to clearly identify individuals (Iannarelli, 1989; Lammi, 2004). Identification includes the identity of the person, which takes the picture and compare to the biometric records available in the database. The biometrics recognition is the advanced technology for finding human being nowadays (Anwar et al., 2015). A human can be identified based on both physiological and behavioral characteristics if it takes following properties like universality, uniqueness, permanent and collectible (Anwar et al., 2015; Hurley et al., 2007).