**Back–end Intelligence component**

The proposed system is designed with the aim of enabling automated surveillance of large infrastructures such as airport, shopping malls, other. Such tasks incorporate massive monitoring of infrastructure visitors in real – time. Monitoring operation is based on an Internet of Things (IoT) installation architecture consisting of: (a)

**Figure 23.** *Reference architecture of the FLYSEC security and safety risk-assessment surveillance system.*

**Figure 21.**

*Deep Learning Applications*

**Figure 22.**

**134**

*FlySec portal - layout visualization.*

*FlySec portal: - intelligent services visualization (upper);* � *automatic passenger classification (lower).*

that act as added value to visitors and simultaneously promote each infrastructure expectations. Indicatively two representative use cases are Navigation services and

*Risk Assessment and Automated Anomaly Detection Using a Deep Learning Architecture*

• Navigation service corresponds to indoor localization and navigation of

information about flight departures or gates status in an airport)

infrastructure and POIs expectations and benefits.

estimation of visitors' luggage abandonment.

techniques (Wi-Fi beacons, NFC tags).

**Data fusion and Risk-based assessment**

**137**

infrastructure visitors in order to assist them in reaching their desired points of interest (POI) not only as quickly as possible but also as efficient and desirable as possible by taking into account user requirements (e.g. disabilities, specific demands) and user location. Moreover, the service provide directions to each visitor via their mobile device to various POIs and informs the user in order to assist them reaching their goal of visit (for example provide information about location of various products in a supermarket, or shops in a mall, or provide

• Recommendation engine aims at providing suggestion of POIs or services that take place inside the infrastructure. The engine takes into account the user profile (information that each user provides optionally during account registration), user feedback (comments, rates), user location and contextual information (time, season, POI status) and create recommendations that are estimated to be assistive to user visiting experience but also promoting

These services aim at monitoring visitors' position and behavior and automatically detect incidents of significant interest such as malicious behavior, anomalous crowd trajectory flow etc. This solution is expected to enhance surveillance procedure for large-scale circumstances where it is demanded in real time, the accurate surveillance of a massive crowd. Indicatively we suggest two surveillance services: Suspicious unattended luggage incidents detection and suspicious visitor loitering detection.

• Unattended luggage incidents detection aims at monitoring in parallel both visitors and the luggage they carry. Such monitoring could be approached either using CCD cameras and approximately detect abandoned luggage for a long period of time, or by tagging luggage (for example using RFID tags) where using RFID scanners in co-operation with visual sensors (CCD cameras)

and human reports (official surveillance personnel), estimate potential

unattended luggage incidents. Moreover, in order to monitor visitors' position, we propose the use of indoor localization techniques using mobile devices in order to have an approximation of visitors' location that willingly allow it, and in addition visual sensors and human reports as well, in order to increase system's awareness of crowd location. Fusion of such information shall be exploited by machine learning algorithms, which result to a coarse grain

• Suspicious loitering detection aims at monitoring visitors' location and in real–

The Data Fusion unit inside the Analytics, Data Fusion and Risk-based Security

time detect anomalous visitors' trajectories or positions that could be suspicious for malicious purposes. Such components may incorporate visual sensors (CCD cameras), human reports and mobile devices localization

Server aims to perform Hard and Soft fusion of heterogeneous data [13–15]

Recommendation engine.

*DOI: http://dx.doi.org/10.5772/intechopen.96209*

**Surveillance services**

**Figure 24.** *Back-end intelligence system architecture.*

various types of sensing devices such as CCD surveillance cameras, QR/barcode scanners, localization equipment (NFC tags, WiFi beacons, etc), RFID scanners, etc., (b) processing units, both centralized and/or distributed, and (c) terminal devices such as mobile phones/tablets, computers, screens, electric signs, etc. (**Figure 24**).

Each sensor device may pre-process the acquired raw data (distributed processing) and the results are gathered on a central cloud-computing infrastructure consisting of independent but co-operative intelligent component each one dedicated for processing data and producing a specific intelligent response for the system. Moreover, the output is transferred to terminal devices. This processing procedure consists of the following steps (**Figure 25**).

The intelligent services are also responsible for automating the monitoring procedure and enhancing visitors' experience. Therefore, we propose two types of services: (a) *Assistance services* and (b) *Surveillance services.*
