Palmprint recognition is a technology that uses the unique composite texture information of the palm surface for automatic identification. In this chapter, first, the main research contents of touchless palmprint recognition are introduced. Second, the establishment of the CUHKSZ large-scale touchless palmprint dataset will be described in detail. Third, parameter optimization experiments are conducted on the CUHKSZ dataset to observe the scientific problems caused by the increase in data scale. Fourth, the experiment comparing the EER of the most used touchless palmprint recognition algorithms on the large-scale dataset will be performed to provide baselines for future research. Finally, the recognition performances of the three most widely used biometric modals, including palmprint, face, and fingerprint, will be compared on the CUHKSZ dataset, which can demonstrate the advantages of touchless palmprint recognition. In the end of this chapter, future research directions will be put forward.
TopIntroduction
Biometric recognition refers to using the inherent physiological or behavioral characteristics of the human body to perform personal identification. In general, physiological characteristics include palmprint, palm vein, palm dorsal vein, fingerprint, face, iris, retina, ear, knuckle print, lip print, voice print, etc., while behavioral characteristics include gait, signature, keyboard typing, and so on. With decades of development, biometric recognition has been widely used in everyday life.
Palmprint recognition is a member of the biometric recognition family. It is a technology that uses the unique features of the palm surface for biometric identification. As shown in Figure 1, the palm contains a wealth of features such as palm shape, principal lines, wrinkles, ridges, minutiae, textures, subcutaneous palm vein, and three-dimensional (3-D) surface curvatures.
Figure 1. Sample images of the palmprint and its features
Advantages of Palmprint Recognition
Among biometric recognition methods, face recognition has problems in situations, such as covering with a mask or goggles, and similar faces between identical twins; fingerprint recognition has issues of counterfeits, wet/dry fingers, and workers and elders who cannot offer clear fingerprints because of years of manual labor or problematic skins. Compared with these recognition methods, palmprint recognition has the advantages of high accuracy, high anti-counterfeiting capability, low privacy sensitivity, and low risk of germ transmission when considering public health, especially during the global COVID-19 pandemic. The prominent palmprint recognition with varieties of advantages has attracted a wide range of attention from academia and industry in recent years (Fei et al., 2018; Zhong et al., 2019).
Categories of Palmprint Recognition
According to different criteria, palmprint recognition can be divided into different categories. Considering the palmprint image dimensions, it can be divided into two-dimensional (2-D) and 3-D palmprint recognition. When taking image resolution as a criterion, it can be divided into high-resolution and low-resolution categories (Zhang & Shu, 1999). According to whether the hand touches the capture device or not, it can be divided into touch-based and touchless palmprint recognition (TLPR). Along with the practical requirements, TLPR is more convenient and flexible. With more attention being paid to it, numerous kinds of TLPR systems have been proposed (Genovese et al., 2014; Zhang et al., 2017; Liang, Guo et al., 2021; Liang, Lu et al., 2022) and it has become the cutting-edge subject of palmprint recognition.