CONTACTLESS AND LESS-CONSTRAINED PALMPRINT RECOGNITION

Biometric systems consist in the combination of devices, algorithms, and procedures used to recognize the individuals based on the characteristics, physical or behavioral, of their persons. These characteristics are called biometric traits. Nowadays, biometric technologies are becoming more and more widespread, and many people use biometric systems daily. However, in some cases the procedures used for the collection of the biometric traits need the cooperation of the user, controlled environments, illuminations perceived as unpleasant, too strong, or harmful, or the contact of the body with a sensor. For these reasons, techniques for the contactless and less-constrained biometric recognition are being researched, in order to increase the usability and social acceptance of biometric systems, and increase the fields of application of biometric technologies. In this context, the palmprint is a biometric trait whose acquisition is generally well accepted by the users. Moreover, palmprints can be captured using low-cost devices, and even in the case of elder people or manual workers. However, biometric systems based on the palmprint traditionally use contact-based acquisitions, with pegs used to constrain the position of the hand in a specific way. For these reasons, this thesis has the objective of researching innovative methods for the contactless and less-constrained recognition of the palmprint. In particular, the researched methods allow to recognize the individuals without the contact of the hand with any surface, and a metric three-dimensional representation is used to eliminate the need for the user to place his hand in a specific position. The originality of the researched techniques allow to perform an accurate biometric recognition, with a focus on the usability, computational speed, and social acceptance of the system. Moreover, the cost of the final device is also taken into consideration. The novelty of the described method, with respect to similar methods in the literature based on contactless three-dimensional acquisitions, resides in the use of an innovative setup, which has a lower cost and captures the images faster. In particular, innovative multiple view acquisition systems, based on CCD cameras and a led illumination, are designed in order to capture the palmprint samples, and original image processing algorithms are implemented to process the samples. Three-dimensional reconstruction techniques are used in order to achieve a metric representation of the hand, invariant to the pose and orientation. Then, pattern recognition methods are implemented in order to extract and match the distinctive features of the palmprints. The novel methods researched in this thesis obtained a good recognition accuracy, in many cases superior to the most recent approaches in the literature. Moreover, good results were obtained regarding the computational speed, usability and social acceptance of the considered methods.