Palmprint identification using restricted fusion
Introduction
In information and networked society, automatic personal identification is an impending and crucial problem that needs to be solved properly. As an efficient and safe solution, biometrics technology has recently been receiving wide attention from researchers. Similar to fingerprint or iris based personal identification, palmprint based identification system can also achieve good performance. For example, it can obtain high accurate recognition rates, and fast processing speed, etc. [1]. At the same time, palmprint based identification system has several special advantages such as stable line features, rich texture features, low-resolution imaging, low-cost capturing devices, easy self-positioning, and user-friendly interface, etc. For these reasons, nowadays the research related to this issue is becoming more active.
So far, there have been many approaches proposed for palmprint recognition including verification and identification, which can be divided into five categories:
(1) Texture-based approaches have been studied extensively, which have shown good performance in terms of recognition rates and processing speed. Zhang et al. proposed PalmCode method, which employed one 2D Gabor filter to extract the texture feature of palmprint [1]. Later, Kong et al. proposed FusionCode using feature-level fusion strategy, which can be regarded as the improved version of the PalmCode [2]. Chen and Xie used dual-tree complex wavelets to extract texture energies of palmprint, and adopted SVM for classification [3]. Li and You proposed a texture-based palmprint retrieval scheme using a layered search strategy for personal identification [4].
(2) Line based approaches are also important for palmprint recognition since lines are essential and basic features of palmprint. Zhang and Zhang used over-complete wavelet expansion and directional context modeling technique to extract principal lines-like features [5]. Han et al. proposed using Sobel’s and morphological operations to extract the line-like features from palmprint images [6]. Lin et al. applied the hierarchical decomposition mechanism to extract principal palmprint features, which includes directional and multi-resolution decompositions [7]. Wu et al. regarded the palm lines as a kind of roof edges, and extracted them according to the zero-cross points of image’s first-order derivative and the magnitude of the edge points’ second derivative [8], [9]. Liu and Zhang extracted the palm lines as a kind of wide lines [10], [11]. Huang et al. proposed a modified finite Radon transformation to extract principal lines [12].
(3) The research on appearance methods is an active area in pattern recognition field. And this technique was also applied to palmprint recognition. Lu et al. [13] and Wu et al. [14] proposed two methods based on principal components analysis (PCA) and linear discriminant analysis (LDA), respectively. Connie et al. proposed several PCA/LDA/ICA-based approaches. And in order to analyze the palmprint images in a multi-resolution-multi-frequency representation, they also adopted wavelet transformation at the same time [15]. Moreover, Hu et al. employed 2D Locality Preserving Projections (2DLPP) for palmprint recognition [16]. Additionally, the Unsupervised Discriminant Projection (UDP) proposed by Yang, which was one of the variations of discriminant LPP, achieved satisfying recognition results [17].
(4) Orientation based approaches are deemed to have the best performance in palmprint recognition field, because orientation feature contains more discriminative information than other features, and is insensitive to illumination changes. Kong and Zhang were the first authors who investigated the orientation information of the palm lines for palmprint verification, which was defined as Competitive Code [18]. Wu et al. proposed another orientation feature based approach named as POC using several directional filters to define the orientation of each pixel [19]. In addition, Ordinal Code proposed by Sun et al. and Robust Line Orientation Code proposed by Jia were also using orientation feature for palmprint recognition [20], [21].
(5) Recently, more and more multi-feature and multi-model based approaches using information fusion technologies have been proposed, since these approaches could provide more reliable results using more features. Kumar and Zhang proposed a multi-feature based approach, which exploited three features, i.e. Gabor based texture feature, line feature and PCA feature [22]. Ribaric and Fratric proposed a biometric identification system based on eigenpalm and eigenfinger features, which extracted the PCA feature from both palm and finger ROI images [23]. Kumar and Zhang proposed an approach combining hand shape and palmprint texture features [24]. Savic and Pavesic presented a multi-model system based on feature extracted from a set of 14 geometrical parameters of the hand, the palmprint, four digitprints and four fingerprints, which obtained very small equal error rate (EER) [25]. In [26], the knuckle-print was integrated a multi-model based scheme. Wang et al. investigated a novel approach exploited palmprint and palm vein using feature-level fusion [27]. Jing et al. proposed a fusion method combining face and palmprint in pixel level and used kernel DCV-RBF classifier for small sample biometrics recognition [28]. Zhao and Huang [29] and Huang et al. [30], [31], [32], [33] proposed a recognition method using modular neural networks [30], [31], [32], [33].
In this paper, we also propose a multi-feature based approach combing principal lines and LPP features for palmprint identification. As we know, the principal lines are the most important and stable feature, which can reflect the whole structure of palmprint. However, only using principal lines for palmprint matching will lose other useful information. On the other hand, LPP feature can be regarded as the statistic feature of palmprint. Thus, in our approach these two features can be jointly used to make a more reliable recognition scheme by exploiting their each advantage.
In this paper, all palmprint images come from Hong Kong Polytechnic University Palmprint Database, which were captured by a CCD-based device described in [1]. This paper is organized as follows. Section 2presents the feature extraction methods. Section 3 describes the matching methods and fusion strategies. Section 4 reports the experimental results. Section 5 offers our concluding remarks.
Section snippets
Extracting principal lines using modified finite Radon transform
In a palmprint image, a palm line can be regarded as a straight line approximately in a small local area. Here, we design a modified finite Radon transform (MFRAT) to extract line feature of palmprint, which is defined as follows [12].
Denoting Zp = {0, 1, … , p − 1}, where p is a positive integer, the MFRAT of real function f[x, y] on the finite grid is defined aswhere C is a scalar to control the scale of r[Lk], and Lk denotes the set of points that make up a line
Identification based on restricted fusion
Identification is the one to many matching. In this section, the matching methods and fusion strategy are introduced as follows.
Experimental results
The proposed method in this paper was tested on the Hong Kong Polytechnic University (PolyU) Palmprint Database. The PolyU Palmprint Database contains 600 grayscale images in BMP image format, corresponding to 100 different palms. In this database, six samples from each of these palms were collected in two sessions, where three samples were captured in the first session and the second session, respectively. The average interval between the first and the second collection was 2 months. The
Conclusions
In this paper, we propose two fusion schemes for palmprint identification exploiting principal lines and LPP features. The principal lines can reflect the whole structure of palmprint. And LPP feature can be regarded as the statistic feature of palmprint. Thus, these two features are jointly used to make more reliable identification schemes by exploiting their each advantage. The results of experiments conducted on PolyU palmprint database show that the proposed schemes can achieve 100%
Acknowledgements
This work was supported by the grants of the National Science Foundation of China, Nos. 60705007 and 60772130, the grant from the National Basic Research Program of China (973 Program), No. 2007CB311002, the grants from the National High Technology Research and Development Program of China (863 Program), Nos. 2007AA01Z167 and 2006AA02Z309, the grant of the Guide Project of Innovative Base of Chinese Academy of Sciences (CAS), No. KSCX1-YW-R-30, and the grant of Oversea Outstanding Scholars Fund
References (34)
- et al.
Palmprint identification using feature-level fusion
Pattern Recognit.
(2006) - et al.
Pattern recognition with SVM and dual-tree complex wavelets
Image Vis. Comput.
(2007) - et al.
Personal authentication using palmprint features
Pattern Recognit.
(2003) - et al.
Palmprint verification using hierarchical decomposition
Pattern Recognit.
(2005) - et al.
Palmprint classification using principal lines
Pattern Recognit.
(2004) - et al.
Palmprint recognition using eigenpalms features
Pattern Recognit. Lett.
(2003) - et al.
Fisherpalms based palmprint recognition
Pattern Recognit. Lett.
(2003) - et al.
An automated palmprint recognition system
Image Vis. Comput.
(2005) - et al.
Two-dimensional locality preserving projecting (2DLPP) with its application to palmprint recognition
Pattern Recognit.
(2007) - et al.
Personal authentication using multiple palmprint representation
Pattern Recognit.
(2005)
Personal recognition based on an image of palmar surface of the hand
Pattern Recognit.
Face and palmprint pixel level fusion and kernel DCV-RBF classifier for small sample biometrics recognition
Pattern Recognit.
Determining the centers of radial basis probabilities neural networks by recursive orthogonal least square algorithms
Appl. Math. Comput.
Palmprint recognition with 2DPCA + PCA based on modular neural networks
Neurocomputing
Online palmprint identification
IEEE Trans. Pattern. Anal. Mach. Intell.
Texture-based palmprint retrieval using a layered search scheme for personal identification
IEEE Trans. Multimedia
Characterization of palmprints by wavelet signatures via directional context modeling
IEEE Trans. Syst. Man Cybernet., Part B
Cited by (38)
A secure image encryption algorithm based on hyper-chaotic and bit-level permutation
2023, Expert Systems with ApplicationsCitation Excerpt :With the development of science and technology, information security plays an increasingly important role in many fields, especially the protection of information about business and personal privacy. In terms of protecting image security, many scholars have proposed plenty of encryption algorithms (Huang, Cheng, Gong, & Zhou, 2020; Kumar, Sheridan, & Bhaduri, 2018; Song, Qiao, & Zhang, 2013; Wang, Quan, & Tay, 2015; Wei, Wenwen, & Li, 2019; Xingyuan, Junjian, & Guanghui, 2019; Zhang & Wei, 2020; Zhou, Fan, Fan, & Li, 2020), and a great deal of the latest technologies can be used in encryption systems (Banan, Nasiri, & Taheri-Garavand, 2020; Fan, Xu, Wu, Zheng, & Tao, 2020; Huang & wing Chau, 2008; Jia, Ling, Chau, & Heutte, 2008; Wang, Xu, Chau, & Xu, 2020; Zhang & Wing Chau, 2009), such as optical transform (Wei & Li, 2019; Wei, Ran, & Li, 2011; Wei, Wang, & Li, 2016; Zhou, Wang, & Gong, 2011), DNA operation (Babaei, Motameni, & Enayatifar, 2019; Nezhad, Safdarian, & Zadeh, 2020; Wang & Li, 2020), chaos (Rakheja, Vig, & Singh, 2020; Shao et al., 2020; Wang & Xu, 2014), bit level permutation (Wang & Wang, 2020; Zhu, Zhang, Wong, & Yu, 2011), sompressed sensing (Zhang, Wang, Sun, & Wang, 2020) and so on. Hyper-chaotic systems have many excellent properties, high sensitivity to initial conditions and control parameters, and have stronger randomness than general chaotic systems.
A level set method based on additive bias correction for image segmentation
2021, Expert Systems with ApplicationsCitation Excerpt :Image segmentation technology is a significant research content in the fields of computer vision, artificial intelligence and image processing (Banan, Nasiri, & Taheri-Garavand, 2020; Fan, Xu, Wu, Zheng, & Tao, 2020; Huang & Chau, 2008; Jia, Ling, Chau, & Heutte, 2008; Cai, Liu, Zhou, Sun, & Li, 2018; Golam Moazzam, 2013; Chan et al., 2018).
Runtime analysis of a multi-objective evolutionary algorithm for obtaining finite approximations of Pareto fronts
2014, Information SciencesCitation Excerpt :Recently, various soft computing techniques have been widely utilized in the fields of science and engineering [30,37,9,22,36,39].
Exploiting the self-organizing financial stability map
2013, Engineering Applications of Artificial IntelligenceCitation Excerpt :Machine learning and artificial intelligence techniques are common means for prediction, modeling and intelligence augmentation in a wide range of engineering applications (e.g., Cheng et al., 2008; Fuertes et al., 2010; Jia et al., 2008; Lin et al., 2006; Muttil and Chau, 2007; Xie et al., 2006; Zhang and Chau, 2009).
Kernel based nonlinear fuzzy regression model
2013, Engineering Applications of Artificial IntelligenceCitation Excerpt :Therefore, such real-life situations are quite often outside the scope of the classical regression analysis (Bargiela et al., 2007), and the classical regression should be extended to deal with fuzzy data. Recent years emerge a surge of literatures about real applications of various artificial intelligence techniques in different engineering fields (without claiming of completeness, see Refs. (Chau, 2006; Muttil and Chau, 2007; Cheng et al., 2002; Lin et al., 2006; Jia et al., 2008; Xie et al., 2006). In this paper, we focus on the fuzzy regression topic as well as its applications.
Incorporation of database approach into contractual issues: Methodology and practical guide for organizations
2013, Automation in Construction