Multi-biometric systems integrating fixed cameras, mobile devices, and drones

Abstract

To identify a person by means of fully automatic systems is still an open problem and a matter of social concern. Many of the approaches coping with this problem are based on the combination of biometrics and computer vision techniques. In particular, biometrics was born in the wake of the wider field of pattern recognition, as that discipline analyses the characteristics of specific individuals with the aim to link their identity to what they are, rather than to what they know or possess. Although much progress has been made in this area, there are still many open problems, which limit its application in daily life at a very large scale. As with many problems in the field of pattern recognition, also for biometrics, the availability of annotated and structured data necessary for designing and validating systems represents a crucial aspect. This theme becomes even more central when systems under consideration show a complex architecture and involve advanced technology such as drones. An in-depth study of the state of the art in this direction has allowed us to identify the most interesting datasets that are currently available. The first goal of this work was to design, acquire, and annotate a large collection of data from different fixed and mobile devices. Data were collected by means of mobile devices such as commercial drones and smartphones, in combination with fixed cameras usually adopted in controlled environments. This type of architecture allows for greater versatility in capturing subjects such as shooting from multiple angles, extreme framing, and using different devices at the same time. Moreover, the characteristics and potential use of this new dataset are drawn. Secondly, we proceeded to design and develop biometric solutions that could demonstrate new integrated approaches of people face trait acquisition. In more details to demonstrate the applicability of drones in a real environment, an acquisition system via a monocular camera installed on-board of a drone has been proposed. This system shows the peculiarity that the drone moves autonomously without a pilot around a cooperative subject (autonomous unmanned drone). The fusion of data acquired by the camera from various perspectives allows us to obtain high-quality aggregate data, useful to be compared with other data obtained from acquisitions made with other devices and protocols. At the end of the acquisition, a 3D face model is obtained by a completely automatic data processing pipeline. A further example of the effectiveness of drones in the biometric field is provided. It is explained the architecture and the results that can be obtained by a drone in building a 3D face model showing a quality comparable to that obtained with a smartphone. Current trends, implications, solutions, and main shortcomings of biometric data protection are discusses. Additionally, a sample study conducted on the combined use of biometrics and cryptography to secure biometric entities is explained and demonstrated. The potential use of these results is addressed, and discusses new advanced methods and applications of biometrics in virtual environments. Finally, the potential uses of these results are addressed, and new advanced methods and applications of biometrics in virtual environments are discussed. Conclusions are dealt with and summarized in the main contributions of the work and provides an insight on future trends in the use of drones in the field of biometrics and in the new era. [edited by Author]