By Frost & Sullivan's Technical Insights Group

With the increasing concerns about terrorist activities, there is a need for reliable technologies to monitor and protect critical infrastructures. Within the past few years, various government agencies and organizations have taken initiatives to protect critical infrastructure assets as well as places of mass gathering (for example, large scale sport events and important international meetings). The existing security models are not scalable and are not suitable for understanding the behavioral patterns of complex systems. Research efforts should focus on devising a reliable security architecture considering inter- and intra-dependencies, cascading effects, while distributing intelligence across the infrastructure to construct a robust security solution.

Addressing the limitations, researchers from University of Genoa (Italy) and Selex Communications S.p.A (Italy) have demonstrated a multi-sensor cognitive system—combination of heterogeneous sensing devices that would function as cognitive radio sensors to enable monitoring and communication including a smart camera network for interpreting interactions and events. SELEX Communications (Finmeccanica group) is a global supplier of secure communication systems designed to protect communities, the environment and critical infrastructures. The cognitive concept facilitates the sensor system to react to situations apart from monitoring the situation. A multi-sensor cognitive-based surveillance system enables automatic interpretation of scenes; and based on the information provided by the sensors it interprets actions and interactions of the observed objects. The researchers have analyzed the impact of computer vision for infrastructure protection and distributed processing and communication techniques on third generation surveillance systems (3GSS) for demonstration purposes.

The model that describes the cognitive system’s behavior is termed as cognitive cycle and it carries these functionalities—sensing, analysis, decision-making and action based on other three characteristics. The sensing system would gather knowledge of the interacting object continuously. The learning process which allows one to have a memory of the past and to anticipate the future through a previous experience consists in the building of the “autobiographical memory.”

“Dynamic Bayesian networks with autobiographical memories have been exploited to understand interactions between objects and are used to specify the topological structure of interactions between the objects in the network,” said Carlo S. Regazzoni, professor at the department of biophysical and electronic engineering, University of Genoa. A cognitive-based approach aims at providing a system with capabilities of direct action on the environment or communication with the necessary personnel for preventive/corrective measures.

The architecture of a cognitive surveillance system comprises sensors (infrared sensors, cameras, radio signal analyzers employing Bluetooth and wireless local area network), a server and processing unit, and actuators (doors and windows locks, palms, speakers connected using wired or wireless methods). The heterogeneous surveillance system characterized by a network of smart cameras and signal analyzer sensors combines two different aspects of intelligence—ambient intelligence (AmI) and cognitive radio (CR). Among other features, AmI allows interaction with cooperative operators and the environment, while CR enables the capability of coping with unexpected or unforeseen situations. “Compared to existing systems, their system can be tuned to analyze and manage critical situations enabling the operators to take appropriate decisions,” Regazzoni added. The architecture can be configured into two levels depending on the interactions performed by the system—interactions with a cooperative operator such as a guardian, and interaction with the environment through an operator.

Figure 1 depicts the architecture of a smart multi-sensor surveillance system. To demonstrate the capabilities of the designed architecture (simulation has been conducted for investigating the integrated information from video and radio sensors), an outdoor scenario as a parking lot in the university campus was considered. The area was monitored using cameras, and cognitive radio sensors. The area was segmented into zones, and it was labeled. The area was thus considered as a graph with nodes as zones and based on the distance between zones, appropriate weights were assigned. It is intended to detect intrusion in a restricted area to identify a person who is involved in jamming action of existing communication network. Based on the information from smart cameras and cognitive radio sensors, the autobiographical memory is trained to predict the intruder’s behavior and to guide the guard (anticipating the intruder’s movement).

According to Regazzoni, some of the key features of the cognitive surveillance systems over conventional security systems for critical infrastructure protection include multi-sensor competency, ability to model interactions and decision-making and actions. The researchers are interested in expanding the capabilities of the multi-sensing system and in joint development programs. They are interested in integrating low-level signal processing and data fusion networks into the cognitive tools, designing tools to detect interactions in a more robust and efficient way, and developing networks to increase the activities. Some focus areas would involve developing a software to analyze directions between different objects in a scene, leverage the additional elements of existing library (utilize libraries from an Italy-based company), and interaction analysis—observe and learn how people and different objects interact in a given scene that can provide additional information as opposed to observing a single object. Products based on this approach are expected in the longer term.

Click to read Frost & Sullivan’s entire report Homeland Security Alert report.