Multidisciplinary Threat Recognition in Homeland Protection Systems

Abstract

The use of a multisensor system, composed of a set of heterogeneous sensors and other devices has already been demonstrated to improve sensibly the recognition capability, through the exploitation of its spatial/capability diversity, given by the presence of multiple devices and coordinated processes which perform threat detection/recognition. In this paper, we evaluate the performance of a multidisciplinary system, which uses a combination of a multisensory classification algorithm and a multidisciplinary fusion rule. This fusion rule combines the decisions coming from different channels with the reasoning process of a machine learning/human in the loop agent. The multidisciplinary fusion rule takes into account the different channel decisions, taken by different sensors and/or devices, and the intelligence provided by the machine learning/ human in the loop channel. The purpose of this channel is to highlight the channels which, inside the machine learning process and through the interaction with the human in the loop agent, show better performance in terms of recognition capabilities in the specific scenario. The performance evaluation of the multidisciplinary threat recognition system is carried out by considering different case studies. The evaluation demonstrates that a multidisciplinary system can classify different threats, by using a set of methods and algorithms, with a high probability of correct classification.

Keywords

• machine learning
• human in the loop agent
• threat recognition
• homeland protection
• data fusion